Cognitive Functional Therapy for Chronic Low Back Pain: A Systematic Review and Meta-Analysis

Abstract

Objective

The objective was to investigate the effectiveness of cognitive functional therapy (CFT) in the management of people with chronic nonspecific low back pain (LBP) and explore the variability in available trials to understand the factors which may affect the effectiveness of the intervention.

Methods

A systematic review with meta-analyses was conducted. Four databases were searched from inception to October 12, 2023. Randomized controlled trials investigating CFT compared with any control group in patients with nonspecific LBP were included. Mean difference and 95% CIs were calculated for pain, disability, and pain self-efficacy. Certainty of evidence was evaluated with the Grading of Recommendations Assessment, Development and Evaluation approach.

Results

Seven trials were included. Low to moderate certainty of evidence was found that CFT was effective for disability at short, medium, and long term time points compared with alternate treatments, including usual care. Low to moderate certainty of evidence was found that CFT is effective for pain in the short and medium terms and probably in the long term. There was high certainty evidence CFT was effective in increasing pain self-efficacy in the medium and long terms. A single study found CFT was cost-effective compared with usual care. Variability was found in the training and implementation of CFT across the included trials, which may contribute to some heterogeneity in the results.

Conclusion

The results show promise in the use of CFT as an intervention likely to effectively manage disability, pain, and self-efficacy in people with chronic nonspecific LBP. The number of clinicians trained, their experience, and quality of training (including competency assessment) may be important in achieving optimal effectiveness.

Impact

This is the most comprehensive review of CFT to date and included investigation of between-trial differences. CFT is a promising intervention for chronic LBP and high-quality synthesis of evidence of its effectiveness is important for its clinical application.

Introduction

Low back pain (LBP) is one of the most prevalent and burdensome health conditions, and the leading cause of global disability.¹ In Australia, LBP contributes significantly to health care costs associated with diagnostics, treatment, and prevention.² Recent research has shown that LBP is a complex, multifactorial condition with physical, psychosocial, and lifestyle risk factors.³ However, many of the commonly used and recommended interventions fail to address the range of factors contributing to each person’s pain and associated disability. Not surprisingly, most current interventions produce only small and short-term effects.⁴

Cognitive functional therapy (CFT) is an integrated method for delivering biopsychosocial management to people with nonspecific LBP, which aims to individualize treatment and coach them to self-manage their condition.⁵ The intervention targets pain-related cognitions, emotions, self-efficacy, and behaviors that are believed to contribute to an individual’s ongoing pain and disability, based on a thorough interview, examination, and clinical reasoning process.

As a relatively new intervention method, showing some promise in early trials,^6,7 there is a quickly growing and changing body of evidence regarding the clinical effectiveness and cost effectiveness of CFT. Existing systematic reviews^8–11 have been limited by a small number of available trials and associated low participant numbers. All existing reviews have important limitations, such as not including recent randomized controlled trials (RCTs)^9,10,12 and including studies that we do not consider to be of CFT.^11,12 With the recent publications, including of the largest CFT trial to date (RESTORE),¹³ it is timely to perform a high-quality synthesis of this evolving body of evidence.

The effects and the duration of effects across some existing CFT trials appear to vary.^13,14 Exploring the characteristics of different trials of CFT may provide insight into potential contributors to this variability; however, previous reviews have not explored this in detail. Given the complexity of the intervention, and potential difficulties in delivering it with high fidelity, factors such as clinician training, intervention dose, and study design issues (such as risk of contamination) may provide useful insights into how to improve future trials and clinical delivery of CFT.

The aims of the current review were to investigate the effectiveness of CFT in the management of nonspecific LBP and to explore the variability in available trials to understand the factors that may have an impact on the effectiveness of the intervention.

Methods

This systematic review was prospectively registered with PROSPERO (CRD42022354909) and adhered to the Preferred Reporting Items Systematic Reviews and Meta-Analysis (PRISMA) guidelines.¹⁵

Literature Search

A comprehensive search was conducted of the following databases: MEDLINE, EMBASE, Cumulative Index to Nursing and Allied Health Literature (CINAHL), and The Cochrane Central Register of Controlled Trials (CENTRAL). A sensitive search strategy was used based on the recommendations of the Cochrane Back and Neck Group for “randomized controlled trials” and “low back pain,”¹⁶ combined with search terms for “Cognitive Functional Therapy”. The full search strategy is included in Supplementary Appendix 1. Development of the search strategy was overseen by a medical librarian and included each database from inception to October 12, 2023. No restrictions were placed on language or setting. The reference lists of included studies and relevant systematic reviews were manually searched for potential studies that were not identified. We also conducted forward citation searching of included articles.

Study Selection and Screening

To be included in this review, studies needed to meet the following eligibility criteria:

1. Population: studies including populations with nonspecific LBP of any duration, with or without associated leg pain. This could include radicular pain or radiculopathy. Nonspecific LBP was defined as pain or discomfort from the lower margin of the 12th rib to the lower gluteal folds. Studies including patients with specific causes of LBP such as cancer, fracture, or infection were excluded.

2. Intervention: studies that investigated the effectiveness of CFT. The intervention must have included individualized assessment and treatment, targeting factors believed to be associated with the individuals’ symptoms. To be considered CFT, the intervention needed to include making sense of pain, exposure with control, and lifestyle change, all individualized to the patient.⁵

3. Comparison: Studies comparing CFT to any control (including but not limited to alternate treatments, usual care, placebo, or no care).

4. Outcomes: Studies needed to report at least 1 of the following outcomes; pain, disability/function, self-efficacy, or economic evaluation.

5. Design: We included RCTs of any design (eg, parallel, cross-over, factorial) and controlled clinical trials using a quasi-randomized method of allocation, such as by alternation or date of birth.

Following conduct of the search, all papers were imported to the Endnote reference management software (Clarivate PLC, London, UK) and duplicates were removed. Two reviewers (L.T. and N.P.) independently screened the title and abstract of each citation. Editorials, commentaries, and conference abstracts were excluded. For each potentially eligible study, 2 reviewers (L.T. and N.P.) examined the full-text article and assessed whether the study fulfilled the inclusion criteria. In cases of disagreement, a decision was made by consensus or, if necessary, a third reviewer was consulted (either P.K. or M.J.H.).

Data Extraction

Data for each included trial were extracted independently by 2 reviewers (L.T. and N.P.) using a standardized piloted data extraction form in an Excel spreadsheet (Microsoft Corp, Redmond, WA, USA). Discrepancies were resolved through discussion. Extracted data included study design (eg, population, sample size, setting), description of interventions (eg, type of intervention, dosage, and level of supervision), effectiveness outcomes of interest, and corresponding follow-up time points.

Timing of Outcome Assessment

Outcome assessment data were extracted for 3 predefined time periods, short-term follow-up (0 to <3 months following randomization), medium-term follow-up (3 to <12 months following randomization), and long-term follow-up (≥12 months following randomization). In studies reporting multiple follow-up periods within the same time-window, we used the period closest to 6 weeks for the short-term, closest to 6 months for the medium-term, and closest to 12 months for the long-term follow-up. We considered the long-term outcome the primary time point for this review.

Methodological Quality Appraisal and Quality of Evidence Assessment

Assessing for Risk/Bias

Risk of bias was assessed independently by 2 authors (P.K. and M.J.H.) using the Cochrane Risk of Bias 2 tool (RoB 2.0). Domains assessed included bias arising from the randomization process, bias due to deviations from the intended interventions, bias due to missing outcome data, bias in the measurement of the outcome, and bias in the selection of the reported result. Any disagreements were resolved by discussion or a third author (L.T.), if required. Further details on how we applied the RoB 2.0 criteria are described in Supplementary Appendix 2.

Assessing Certainty of Evidence

The overall quality of evidence was assessed for each contrast and outcome using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system.¹⁷ The GRADE classification was downgraded by 1 level if any of the following were present: methodological quality—when >50% of included participants came from studies rated as low methodological quality; inconsistency of results—based on observation of the variability of point estimates across individual trials and the I² statistic (when only a single trial was available for a given contrast, we automatically downgraded for inconsistency); imprecision—based on inspection of the 95% CI of the pooled estimate to determine if it included values that would have substantially different clinical implications; publication bias—downgraded if a funnel plot suggested the presence of publication bias (only conducted if >10 studies were available for meta-analysis) or if other evidence of publication bias exists such as other evidence of mostly small studies with mostly positive results, industry sponsorship, multiple trials in registries but with no published results; and indirectness—indicated when either the population, intervention, comparison, and outcome was not directly assessed.

Data Analysis

When available, the reported between-group effect sizes adjusted for baseline score of the outcome were used in analyses, as recommended by the Cochrane handbook.¹⁸ If unavailable, raw mean ± SD outcome data were extracted for the intervention group and comparison group at baseline and follow-up time points. All outcomes were converted to a scale from 0 to 100 to enable pooling and easier comparison of effects across outcomes.¹⁹

We performed a random effects meta-analysis for each outcome (pain, disability, and self-efficacy) at each time point if adequate homogeneity existed (in terms of population, intervention type, control, and outcome), using Comprehensive Meta-Analysis Version 4 (Biostat Inc, Englewood, NJ, USA). The primary analysis included studies comparing CFT to all alternate treatments, including usual care. A secondary analysis was then run separately comparing CFT to alternate treatments excluding usual care, and to usual care only. Where an included study had used mixed models and/or imputed the data for their intention to treat analysis, we included the full baseline sample size value in the meta-analysis, because a value for each participant had contributed to the result.

For aim 2, we provide a descriptive narrative synthesis of important characteristics across the included studies that may have an impact on the effectiveness of the intervention. These characteristics included prior clinician experience, CFT training, CFT dosage, population/setting, and trial factors (including comparison treatment group, fidelity testing, and number of clinicians trained) as outlined in Table 1.

Table 1.

Comparison of CFT in Included Trials^a

Category	Sub-Category	Data From the Following Study
		Ahmad et al 22	Avila et al 23	Castro et al 14	Kent et al 13	Ng et al 21	O’Keeffe et al 20	Vibe Fersum et al 7
Prior experience	Treating clinician prior experience (y)	1b	7	Newly graduated	Minimum of 2	15	15 or 16b	6–13b
	CFT experience prior to trial (y)	0.5b	3	NR	<4 d of CFT training	NR	NR	NR
CFT training	Experience of CFT mentor	>5 y of CFT experience	>5 y of CFT experience	>4 y of CFT experience	>10 y of CFT experience	>10 y of CFT experience	>10 y of CFT experience	>10 y of CFT experience
	Amount of CFT training (h)	100b	106	106	Minimum of 80	NR	>80	106
	Did the training involve mentoring while treating patients?	Yesb	Yes	Yes	Yes	Yesb	Yes	Yes
	Was there a competency check?	NR	NR	NR	Yes	Yesb	Yes	Yesb
Dosage of CFT	No. of sessions (mean)	Protocol: 12 Reported: NR	Protocol: 4–12 Reported: 5.6	Protocol: 5 Reported: 4.2	Protocol: 8 Reported: 7	Protocol: 4 Reported: 3.6	Protocol: 4–16 Reported: 5.0	Protocol: 7 or 8 Reported: 7.7
	Length of sessions (h)	1	1	1	0.5–1	0.5–1	0.5–1	0.5–1
	Treatment time period (wk)	8	11	8	12 with single booster session at 26	8	13.7	12
Population/setting	Clinical setting of CFT intervention	Primary care	Primary care	Primary care	Primary care	Primary care	Primary care	Primary care
	Baseline disability of CFT group, converted mean/100 (SD)	39.80 (8.02)	38.05 (12.49)	30.7 (10.82)	55.4 (26.7)	20.4 (21.3)	32.1 (12.6)	21.3 (7.5)
Trial factors	Comparison treatment group	MSI-based intervention	Core exercise and manual therapy	Core exercise and manual therapy	Usual care	Usual care	Group-based exercise and education	Manual therapy and exercise
	Did the same clinicians treat CFT and control group participants?	No	No	No	No	No	Yes	No
	Did clinicians in the same environment treat CFT and control group participants?	No	No	Yes	No	No	No	NR
	No. of clinicians trained for trial	CFT: 1 Control: 1	CFT: 1 Control: 1	CFT: 1 Control: 3	18	1	CFT + control: 3	3
	Was there fidelity testing?	NR	Yes	Yes	Yes	NR	Yes	Yes

^a CFT = cognitive functional therapy; MSI = movement system impairment; NR = not reported.

^b Data gathered from communication with authors.

Results

Our search identified 1949 independent studies. After screening by title and abstract, 117 remained and were assessed by full text, after which 7 RCTs (reported in 8 manuscripts) were included in our review.^{6,7,13,14,20–23} The reasons for exclusion are detailed in Figure 1 and the full search strategy is included in Supplementary Appendix 1. For the 3-arm RESTORE Trial,¹³ we chose to compare 2 of the 3 trial arms—usual care and CFT only—as the third arm (CFT plus wearable sensor biofeedback) was a combination treatment.

Figure 1.

Flow diagram of study selection from 4 databases.³¹ CFT = cognitive functional therapy; LBP = low back pain; RCT = randomized controlled trials.

Table 2 outlines the characteristics of the 7 included trials, with a combined sample size of 1011 individual participants. Six studies included adults with chronic LBP.^{7,13,14,20,22,23} One study met our inclusion criteria, but was excluded from meta-analyses as the population was quite different to the other studies (eg, lower age and only rowing-related pain).²¹ Three studies compared CFT to an intervention including manual therapy and exercise^6,14,23 and 2 studies compared CFT to usual care.^13,21 One study compared CFT to group exercise and education.²⁰ One study compared CFT to movement system impairment (MSI)–based treatment which involves classification into a direction-specific subgroup, education on normal postures and exercises to improve trunk movement.²²

Table 2.

Characteristics of RCTs Included in the Systematic Review^a

Study	Participants	Outcome	Follow-Up (wk)	Intervention and Control	Dosage
Ahmad et al22 (2023)	This study included patients who had LBP (according to the criteria of the International Classification of Diseases, 10th Revision [ICD-10]; website: https://icd.who.int/), classified by a duration of pain of >12 wk (nonspecific LBP), who were > 18 y old, and who were eligible to participate (N = 91) Mean age: CFT group = 26 y (SD = 3.17 y); MSI group = 27.17 y (SD = 5.10 y) No. (%) women: CFT group = 28/45 (62); MSI group = 26/46 (57)	LBP intensity: NRS LBP disability: ODI	8, 26	Intervention: CFT Control: MSI-based intervention	Intervention: 12 one-h sessions over an 8-wk treatment period Control: 11 one-h sessions over an 8-wk treatment period
Avila et al23 (2023)	Participants were included if they were 18–75 y old, were independently mobile (with or without aids), were seeking treatment for LBP with a duration of at least 12 wk after surgical intervention in the lumbar spine for lumbar or sciatic pain, and were able to speak and understand Portuguese; also, in accordance with the criteria of Vibe Fersum et al7 and O’Keeffe et al,20 a score of >14% on the ODI as well as at least 3 points on the NPRS were required for inclusion (N = 80) Mean age: CFT group = 49 y (SD = 10.59 y); CE/MT group = 50.45 y (SD = 10 y) No. (%) women: CFT group = 24/40 (60); CE/MT group = 20/40 (50)	LBP intensity: NRS LBP disability: ODI	After intervention (mean = 11 wk for CFT; mean = 10 wk for CE/MT), 22	Intervention: CFT Control: core exercises and manual therapy; based on McGill core exercises (McGill Curl Up; Side Plank; Bird Dog)	Intervention: weekly sessions of 1 h varied pragmatically from 4 to 12 wk Control: weekly sessions of 1 h varied pragmatically from 4 to 12 wk
Castro et al14 (2022)	Participants were 18–65 y old, were independently mobile (with or without aids), were seeking treatment for LBP with a duration of at least 12 wk, were able to speak and understand Portuguese, and had a score of >14% on the ODI (N = 148) Mean age: CFT group = 46.4 y (SD = 10.6 y); core-MT group = 40.4 y (SD = 11.6 y) No. (%) women: CFT group = 44/74 (59.5); core-MT group = 47/74 (63.5)	LBP intensity: NRS LBP disability: ODI	8, 26, 52	Intervention: CFT with filmed home exercise program Control: core exercises for strengthening and endurance; based on McGill core exercises	Intervention: 5 one-h sessions for an 8-wk intervention period, on a weekly basis for the first 2 wk and then a session every 2 or 3 wk Control: 5 one-h sessions for an 8-wk intervention period, on a weekly basis for the first 2 wk and then a session every 2 or 3 wk
Kent et al13 (2023)	Participants were adults (≥18 y old) who had chronic LBP lasting >3 mo, had sought care from a primary care clinician for their back pain at least 6 wk previously, had an average back pain intensity of 4 or more on an NPRS from 0 to 10, and had at least moderate pain-related interference with normal work or daily activities measured by item 8 of the 36-Item Short-Form Health Survey12 (N = 329) Mean age: UC group = 47.7 y (SD = 16 y); CFT-only group = 47.5 y (SD = 15 y) No. (%) women: UC group = 98/165 (59); CFT-only group = 99/164 (60)	LBP intensity: NPRS LBP disability: RMDQ Pain self-efficacy: PSEQ	6, 13, 26, 52	Intervention: CFT Control: usual care; free to seek treatment from other health care professionals	Intervention: participants received up to 7 treatment sessions over 12 wk plus a “booster” session at 26 wk (initial consultation ~60 min, follow-up sessions ~30–40 min) Control: NA
Ng et al21 (2015)	Participants were adolescent male rowers between 14 and 19 y old, with between 1 and 4 y of school-level rowing experience, and with LBP related to rowing at the time of data collection; the inclusion criteria were participants rowing competitively in local rowing regattas; a self-reported LBP intensity of >3/10 on a visual analog scale, which must be reached during a typical rowing training session; and pain location within the lumbar region as drawn on a diagram (N = 36) Mean age: CFT group = 16.3 y (SD = 1.5 y); UC group = 15.2 y (SD = 1.5 y) No. (%) women: 0 (0)	LBP intensity: NPRS LBP disability: RMDQ	8, 12	Intervention: CFT Control: active control; remained free to seek treatment from health care providers external to the project; the nature of treatment received by the control group from external health care providers was not recorded	Intervention: the initial session was ~1 h in duration, and follow-up appointments were 30 min; rowers were seen 1 wk after the initial session and then fortnightly after that; total duration of the intervention: 8 wk Control: NA
O’Keeffe et al20 (2020)	Patients were eligible for inclusion if they met all of the following criteria: between 18 and 75 y old, having nonspecific chronic LBP for a duration of at least 6 mo, having a score of 14% or more on the ODI, being independently mobile (with or without aids) to be capable of participating in a rehabilitation program, and being able to speak and understand English well enough to complete the questionnaires independently (N = 206) Mean age: CFT group = 47.0 y (SD = 13.2 y); group intervention = 50.6 y (SD = 14.9 y) No. (%) women: CFT group = 82/106 (77.4); group intervention = 70/100 (70)	LBP intensity: NPRS LBP disability: ODI Pain self-efficacy: PSEQ	After intervention (mean = 14 wk for CFT; mean = 4 wk for group intervention), 26, 52	Intervention: CFT Control: group exercise and education	Intervention: the length of the CFT intervention varied in a pragmatic manner based on the clinical progression of the participants Control: up to 6 classes over 6–12 wk, each lasting ~1 h 15 min, with up to 10 participants in each class
Vibe Fersum et al7 (2019)	Patients were eligible for the study if they were between 18 and 65 y old, were diagnosed with nonspecific LBP for >3 mo that was primarily localized from T12 to the gluteal folds, and reported that their pain was provoked and relieved with postures, movement, and activities; pain intensity measured with a pain intensity NRS over the last 14 d of >2/10 and an ODI score of >14% were necessary to be admitted to the study (N = 121) Mean age: CFT group = 41.0 y (SD = 10.3 y); MT-EX group = 42.9 y (SD = 12.5 y) No. (%) women: CFT group = 32/62 (51.6); MT-EX group = 31/59 (52.5)b	LBP intensity: NPRS LBP disability: ODI	12, 52, 156	Intervention: CFT Control: manual therapy and home exercises	Intervention: the initial session was 1 h, and follow-up sessions were 30–45 min; patients were seen on a weekly basis for the first 2 or 3 sessions and then progressed to 1 session every 2–3 wk during the 12-wk intervention period Control: 1-h initial consultation and 30 min for follow-up sessions

^a CE/MT = core exercise and manual therapy; CFT = cognitive functional therapy; core-MT = McGill core exercises and manual therapy; LBP = low back pain; MSI = movement system impairment; MT/EX = manual therapy and exercise; NA = not applicable; NPRS = Numeric Pain Rating Scale; NRS = numeric rating scale; ODI = Oswestry Disability Index; PSEQ = Pain Self-Efficacy Questionnaire; RCTs = randomized controlled trials; RMDQ = Roland-Morris Disability Questionnaire; UC = usual care.

^b Data gathered from correspondence with author.

Primary Analyses: CFT Versus Alternate Intervention Including Usual Care

Below we synthesize the pooled results where at least 2 trials reported on pain, disability, or self-efficacy for the same time window. The complete results, including contrasts where only a single trial was available, are reported in Table 3.

Table 3.

Summary of Results^a

Comparison/Outcome	No. of Participants	No. of Studies	MD (95% CI)	P	Grade
Primary analysis: CFT vs alternative care
Pain intensity
Short term	64813,14,22,23	4	13.8 (5.5 to 22.2)	.001	Low
Medium term	9757,13,14,20,22,23	6	13.7 (9.3 to 18.2)	<.001	Moderate
Long termb	8047,13,14,20	4	7.8 (−0.2 to 15.8)	.06	Low
Disability
Short term	64813,14,22,23	4	10.8 (6.6 to 14.9)	<.001	Low
Medium term	9757,13,14,20,22,23	6	10.6 (6.6 to 14.6)	<.001	Moderate
Long termb	8047,13,14,20	4	8.8 (1.8 to 15.9)	.01	Low
Pain self-efficacy
Short term	32917	1	8.7 (4.0 to 13.2)	<.001	Moderate
Medium term	53513,20	2	12.3 (8.4 to 16.2)	<.001	High
Long termb	53513,20	2	12.5 (8.6 to 16.5)	<.001	High
CFT vs exercise/MT/education
Pain intensity
Short term	31914,22,23	3	14.7 (2.9 to 26.5)	.01	Low
Medium term	6467,14,20,22,23	5	13.4 (7.7 to 19.1)	<.001	Low
Long termb	4757,14,20	3	4.7 (−2.9 to 12.4)	.23	Low
Disability
Short term	31914,22,23	3	9.5 (5.0 to 14.0)	<.001	Moderate
Medium term	6467,14,20,22,23	5	9.1 (5.4 to 12.8)	<.001	Moderate
Long termb	4757,14,20	3	5.3 (1.3 to 9.2)	<.01	Moderate
Pain self-efficacy
Medium term	20620	1	11.1 (4.0 to 18.2)	<.01	Low
Long termb	20620	1	10.0 (2.5 to 17.5)	<.01	Low
CFT vs usual care
Pain intensity
Short term	32913	1	11.0 (6.0 to 16.0)	<.001	Moderate
Medium term	32913	1	14.0 (9.0 to 19.0)	<.001	Moderate
Long termb	32913	1	15.0 (9.0 to 210)	<.001	Moderate
Disability
Short term	32913	1	15.4 (10.2 to 20.6)	<.01	Moderate
Medium term	32913	1	20.0 (14.6 to 25.4)	<.01	Moderate
Long termb	32913	1	20.0 (14.6 to 25)	<.01	Moderate
Pain self-efficacy
Short term	32913	1	8.7 (4.0 to 13.2)	<.001	Moderate
Medium term	32913	1	12.8 (8.2 to 17.5)	<.001	Moderate
Long termb	32913	1	13.5 (8.9 to 18.1)	<.001	Moderate

^a CFT = cognitive functional therapy; GRADE = Grading of Recommendations Assessment, Development and Evaluation; MD = mean difference; MT = manual therapy.

^b Primary time point.

Pain Intensity

Four trials (n = 648)^13,14,22,23 investigated the short-term effects of CFT compared with an alternate treatment (eg, exercise and manual therapy) or usual care for pain intensity. The pooled results provided a low certainty of evidence that CFT was more effective at reducing pain in the short-term follow-up (mean difference [MD] = 13.8; 95% CI = 5.5 to 22.2). Six trials (n = 975)^{7,13,14,20,22,23} investigated medium-term effects and the pooled results provided a moderate certainty of evidence that CFT was more effective (MD = 13.7; 95% CI = 9.3 to 18.2). Four trials (n = 804)^7,13,14,20 investigated long-term effects and provided a low certainty of evidence that CFT is probably effective (MD = 7.8; 95% CI = −0.2 to 15.8) compared with an alternate treatment or usual care. A summary of results is provided in Table 3 and Figure 2.

Figure 2.

Forest plots for all primary outcomes. CFT = cognitive functional therapy; Con = control; Int = intervention.

Disability

Four trials investigated CFT compared with either usual care or alternative treatment for disability in the short term (n = 648).^13,14,22,23 Pooled results provided a low certainty of evidence favoring CFT compared with controls (MD = 10.8; 95% CI = 6.6 to 14.9). Six trials (n = 975)^{7,13,14,20,22,23} investigated the medium-term effect of CFT compared with usual care or alternate treatment. Pooled results provided a moderate certainty of evidence favoring CFT (MD = 10.6; 95% CI = 6.6 to 14.6). Four studies (n = 804)^7,13,14,20 investigated the long-term effect of CFT and found a low certainty of evidence in favor of CFT, compared with other treatments or usual care (MD = 8.9; 95% CI = 1.8 to 15.9). A summary of results is provided in Table 3 and Figure 2.

Pain Self-Efficacy

Two trials (n = 535)^13,20 investigated the medium-term effect of CFT compared with alternative treatment or usual care for pain self-efficacy. Pooled results provided high certainty evidence in favor of CFT (MD = 12.3; 95% CI = 8.4 to 16.2). Two trials (n = 535)^13,20 provided high certainty evidence in favor of CFT when compared with alternative treatment or usual care in the long term (MD = 12.5; 95% CI = 8.6 to 16.5). A summary of results is provided in Table 3 and Figure 2.

Cost Outcomes

Only 1 study reported on cost outcomes. The RESTORE trial used a societal perspective and found that CFT was both more effective and less costly compared with usual care with an incremental gain of 0.12 quality-adjusted life-years per participant (95% CI = 0.08 to 0.16), at a lower overall cost of $AUD −5276 (95% CI = −$10,529 to −$24).¹³

Secondary Analyses

Our secondary analyses separately compared CFT to alternative treatment excluding usual care (ie, manual therapy and exercise, exercise and education, and MSI-based treatment) and CFT to usual care. The results are presented in Table 3 and Supplementary Appendix 3. For the outcome of pain intensity in the short and medium terms (short-term n = 319^14,22,23; medium-term n = 646^{7,14,20,22,23}), we found a low certainty of evidence favoring CFT compared with alternative interventions excluding usual care (short-term MD = 14.7; 95% CI = 2.9 to 26.5; medium-term MD = 13.4; 95% CI = 7.7 to 19.1). We found a low certainty of evidence that CFT may not be superior in the long term (n = 475^7,14,20; MD = 4.7; 95% CI = −2.9 to 12.4).

For the outcome of disability across all time points (short-term n = 319^14,22,23; medium-term n = 646^{7,14,20,22,23}; long-term n = 475^7,14,20), we found a moderate certainty of evidence that CFT was superior to alternative treatment excluding usual care (short-term MD = 9.5; 95% CI = 5.0 to 14.0; medium-term MD = 9.1; 95% CI = 5.4 to 12.8; long-term MD = 5.3; 95% CI = 1.3 to 9.2).

For the secondary analysis of CFT compared with usual care, single outcome results are shown in Table 3 and Supplementary Appendix 3.

Aim 2

Details of the 7 included trials that may affect outcomes are presented and contrasted in Table 1. Clinician training varied in some aspects between trials. Four studies^6,14,22,23 provided 100 to 106 hours of CFT training, while another 2^13,20 reported a minimum of 80 hours of face-to-face training. All 7 of the included studies^{6,13,14,20–23} reported including practice with real patients in the training, with mentoring, supervision, and feedback of this practice. Four of the 7 studies^6,13,20,21 included a clinical competency check before allowing therapists to provide care in the trial. The experience of the CFT trainers varied between trials. In 4 studies,^6,13,20,21 the trainers were either the originators of CFT or had been coached by the originators, while in 3 studies,^14,22,23 the experience and training of the CFT trainers was unclear.

Dosage of CFT varied moderately across the 7 trials. The mean or median number of sessions reported within each trial varied between 3.6 and 7.7 sessions, recognizing that the number of CFT sessions is individualized and therefore may vary across clinical populations. Two trials delivered a mean of 3.9 sessions,^14,21 2 trials delivered between 5 and 6 sessions,^20,23 and 3 trials delivered 7 or more sessions.^7,13,22 Three studies had an intervention period of 8 weeks,^14,21,22 and 3 studies had an intervention period of at least 12 weeks.^7,13,20 One study had a pragmatic duration of 4 to 12 weeks.²³ The RESTORE study was the only study which provided a “top up” session at 6 months.¹³ Across all studies, the duration of each individual session of CFT ranged from 30 to 60 minutes.

Trial factors such as the number of clinicians trained and the risk of contamination varied across trials. One trial¹³ trained 18 physical therapists, 2 trials^7,20 trained 3 physical therapists, and 4 trials^14,21–23 trained just 1 physical therapist to deliver the CFT intervention. The clinical experience of the physical therapists varied from “newly graduated”¹⁴ to experienced physical therapists.¹³ In 1 trial,²⁰ the same physical therapists treated the intervention and control patients and, in another trial,¹⁴ intervention and control patients were treated in the same clinic increasing the chance of contamination. Another trial factor that varied between trials was the control treatment. Two trials used a usual care control,^13,21 3 used manual therapy and exercise,^7,14,23 1 used MSI,²² and 1 used group-based exercise and education intervention.²⁰

Some of the above factors may have influenced the heterogeneity in the effect sizes observed between trials. For example, characteristics that differed between the study with the smallest effect size (Castro et al¹⁴) and the largest effect size (RESTORE),¹³ included CFT trainer experience, number of clinicians trained, prior experience of clinicians, training clinicians to a competency standard, risk of contamination, and control treatment.

Deviations From Protocol

We made 1 adjustment to our published protocol, namely the conversion of all outcomes onto a 0–100 point scale, rather than using standardized mean difference. So, pain intensity (scored from 0 to 10 points), Roland-Morris Disability Questionnaire (RMDQ) (scored from 0 to 24 points), and Pain Self-Efficacy Scale (scored from 0 to 40 points) scores were all converted to a 100-point scale. Oswestry Disability Index (ODI) scores were already reported on a scale from 0 to 100. We did this because presenting results in this form made it easier to contrast results across different outcomes, it improves the clinical interpretability for clinicians,¹⁹ and there is some evidence for the validity of this approach.¹⁹

Discussion

Key Findings

This review found a low to moderate certainty of evidence that CFT is effective in reducing disability when compared with alternate interventions including usual care across all time points. Our findings were more variable for pain intensity, where we found a moderate certainty of evidence that CFT was more effective than alternative interventions including usual care, in the medium term and a low certainty of evidence that it is superior in the short and probably in the long term. There was high certainty evidence that CFT effectively increases pain self-efficacy at medium and long-term time points. A single study investigated cost effectiveness and found CFT was more cost effective than usual care.

Our secondary analysis comparing CFT to alternative interventions only (ie, manual therapy and exercise, group exercise and education, or MSI-based treatment) found a moderate certainty of evidence that CFT was effective in reducing disability across all time points.

We explored differences in the training and implementation of CFT across available trials to identify factors that may contribute to the variability in the observed results. Some elements that varied between trials and seem most likely to contribute to different effects include number of clinicians trained (only 1 clinician in the trial with the smallest effects¹⁴ compared with 18 in the trial with the largest effects¹³), experience of clinicians prior to the trial (<1 year in the trial with the smallest effect¹⁴ and a minimum of 2 years in the trial with the largest effect¹³), risk of contamination (patients in CFT and control groups treated in the same clinic in the trial with the smallest effects¹⁴), and competency checking before clinicians started delivering CFT (no reporting of competency checking in the trial with the smallest effects¹⁴ compared with competency checking reported in the trial with the largest effects¹³). A recent scoping review investigating the training of physical therapists to deliver individualized biopsychosocial interventions suggested key components of training that were necessary included: skilled supervisors providing mentoring and feedback while trainee’s work with patients, training to a competency standard, and the monitoring of treatment fidelity within clinical trials.²⁴ These components align with some of the key differences we found between trials that demonstrated small versus large effects.

Comparison to Previous Literature and Meaning of the Findings

A previous systematic review investigated the CFT for pain and disability outcomes, and concluded the effectiveness of CFT is uncertain with most CIs being very wide.⁸ With the addition of several new and larger studies included in our review we are able to draw more confident and often different conclusions. For example, we found a moderate certainty of evidence that CFT is effective for pain and disability in the medium term compared with the previous review by Devonshire and colleagues that found a very low certainty of evidence of uncertain effect.¹² Their review included 5 studies, only 3 of which were included in our review. The study by Khodadad²⁵ was excluded from our review because CFT was delivered as a group intervention and was deemed contrary to the individualized nature of CFT.⁵ The study by Sheeran²⁶ was excluded from our study because it was a conference abstract only. We included 3 new studies published since the previous review, including an RCT that is substantially larger (in sample size and the number of clinicians providing the treatment) than any study in the previous review.^13,22,23 We also included the study by Ng²¹ that was not included in the synthesis of the previous review, as we did not limit the age of participants in our inclusion criteria. However, we did not include this study in any meta-analyses as the population of young rowers was considered substantially different to other included studies. The total number of participants in our review (n = 1011) is double that of the previous review by Devonshire and colleagues⁸ (n = 507). An earlier systematic review¹⁰ published in 2022 by Miki and colleagues found a very low certainty of evidence that CFT was superior to controls for pain intensity in the long term, and disability in the medium and long term. That review included 3 studies which were included in the current review, however their findings were limited by a small number of available published trials. Zhang and colleagues¹¹ also conducted a review on CFT, which included 2 trials that did not meet our criteria for individualized CFT (Khodadad²⁵ and Elshiwi²⁷). We also included a recent trial which compared CFT to a MSI intervention.²² The present review is the most up to date synthesis of data from clinical trials of CFT.

Our study reports substantially higher certainty evidence (6 of 9 contrasts for our primary outcome have GRADE ratings of moderate or high) (Tab. 3) and greater precision than both previous reviews, due to the additional studies and due to different assessments of individual trial RoB and application of criteria for downgrading GRADE rating. Our criteria for judging RoB and applying GRADE are clearly reported in Supplementary Appendix 2.

Clinical Implications

This review suggests that CFT is probably superior to usual care or alternate interventions, such as manual therapy, MSI, exercise, and education for patients with chronic LBP, and the effects on disability, self-efficacy and probably pain are maintained in the long term. As such, CFT could be considered as an intervention option for people with chronic LBP.

There is uncertainty about what is considered a clinically worthwhile effect size in this population. It has been argued that some important factors to consider are the benefits, harms and inconveniences of a treatment compared with the proposed alternative, and the opinions of the health care recipients, who are to make informed decisions about the care they receive.²⁸ In the present review, the CIs for disability in the long-term outcomes span from 1.8 to 15.9 and therefor include values that are likely to be considered both worthwhile and not worthwhile by some patients. However, rather than predefining clinical significance, we believe these results should be presented to patient’s seeking care, ensuring shared decision-making and truly individualized care. It has been suggested that this more nuanced approach has more clinical relevance than a benchmark effect size.²⁹

All the available trials involved substantial training of the clinicians (>80 hours) so it is unclear how effectively this intervention can be delivered by clinicians with less training, and therefore how widely available CFT is to current patients with chronic LBP. In the study by Castro et al¹⁴ with the smallest effects, the intervention was delivered by a single recently graduated physical therapist who was not formally assessed for competency, limiting the generalizability of the findings. The training was provided by a physical therapist who had also only undergone limited CFT training (attended CFT workshops as an observer). It is possible the smaller effects in this trial are related in part to the training quality limitations, but there may also be other reasons. Determining the critical elements of training to achieve high levels of competence in delivering CFT is a priority for future research. For example, a recent article identified the need for effective communication and shared decision making between clinician and patient to more successfully implement psychologically-informed practice in the management of chronic pain.³⁰

Strengths and Limitations

To our knowledge, this is the largest and most comprehensive systematic review investigating CFT, and the only review specifically aiming to describe factors associated with the individual trials which may have contributed to differences in the results. While Aim 2 was a descriptive exploration of factors which may have an impact on the effectiveness of the intervention, we never intended to perform meta-regression, moderator or mediation analyses as we correctly assumed there would be inadequate data for such analyses. The inclusion of the several new RCTs, including the RESTORE trial,¹³ means there is an increasing number and proportion of participants from higher quality clinical trials, which had been a limitation of earlier reviews.⁸ As expected, there were some differences between the RCTs included in our meta-analyses, but we considered those trials similar enough to pool. We explored differences in effects when the control group was an alternative intervention such as manual therapy and/or exercise, or usual care. Our exploration of differences between trials, such as clinician training and fidelity testing, enables readers to better understand potential factors which may affect individual study results, but we cannot be certain that these were important.

Our primary analysis included trials with either usual care or alternative intervention control groups, and our secondary analyses were conducted separately based on the control interventions. This enables readers to assess the likely effectiveness of CFT compared with both or either of these comparisons, and to also view the variability of effects from individual studies. We included trials that we considered to assess the effectiveness of CFT; however, there are naturally some variations in trial design, with some being less pragmatic than others. We did not include gray literature searches, so we do not know if registered trials have been completed but not published, therefore we cannot exclude that potential publication bias. One author has played a leading role in the development, teaching and research of CFT (P.O.), and 2 authors have been involved in previous research of CFT (P.K. and M.J.H.), which could represent a potential source of bias.

Future Directions

We believe this review provides important guidance for future investigations of CFT as an intervention in the management of nonspecific LBP. To date, there has been important variability in the clinician training and treatment delivery of CFT, including comparison groups, which likely have an impact on the estimated treatment effect. In future, as more studies are available, meta regression could be used to quantify the extent to which those study level differences explain differences in effects between studies.

Conclusion

CFT appears to be effective compared with alternative treatments and/or usual care for improving disability, pain, and self-efficacy in patients with chronic LBP. We identified sources of variability between trials, such as quality of CFT training, competency and fidelity assessment, and the number of clinicians trained. These factors should be investigated in future studies to determine their impact on the clinical effectiveness of CFT.

Author Contributions

Lena Thiveos (Conceptualization, Data curation, Formal analysis, Methodology, Project administration, Writing—original draft, Writing—review & editing), Peter Kent (Conceptualization, Data curation, Formal analysis, Methodology, Project administration, Supervision, Writing—original draft, Writing—review & editing), Natasha Pocovi (Data curation, Methodology, Supervision, Writing—review & editing), Peter O’Sullivan (Conceptualization, Methodology, Writing—review & editing), and Mark Hancock (Conceptualization, Data curation, Formal analysis, Methodology, Project administration, Supervision, Writing—original draft, Writing—review & editing).

Funding

There are no funders to report.

Systematic Review Registration

This systematic review was prospectively registered with PROSPERO (CRD42022354909) and adhered to the Preferred Reporting Items Systematic Reviews and Meta-Analysis (PRISMA) guidelines.¹⁵

Disclosures

Peter Kent and Peter O’Sullivan were coauthors on a publication that reported on the RESTORE trial.¹³

The authors completed the ICMJE Form for Disclosure of Potential Conflicts of Interest and reported no other conflicts of interest.

Data Availability

The data used in this study can be made available by request to the corresponding author.