The mechanisms of effect of a physiotherapist-delivered integrated psychological and exercise intervention for acute whiplash-associated disorders: secondary mediation analysis of a randomized controlled trial.

INTRODUCTION: Integrated psychological and physical treatments can improve recovery for whiplash-associated disorders (WADs). Little is known about how these interventions work.
OBJECTIVE: To examine the mechanisms by which a physiotherapist-delivered integrated intervention for acute WAD improves health outcomes.
METHODS: Secondary analysis using structural equation modelling of a randomized controlled trial comparing integrated stress inoculation training and exercise to exercise alone for acute WAD. Outcomes were disability, pain self-efficacy, pain intensity, and health-related quality of life at 12 months. The intended intervention target and primary mediator, stress was tested in parallel with pain-related coping, an additional cognitive behavioral mediator that significantly improved at posttreatment (Model 1). Stress-related constructs that commonly co-occur with stress and pain were also tested as parallel mediators: depression and pain-related coping (Model 2); and posttraumatic stress and pain-related coping (Model 3).
RESULTS: Reductions in stress mediated the effect of the integrated intervention on disability (β = -0.12, confidence interval [CI] = -0.21 to -0.06), pain self-efficacy (β = 0.09, CI = 0.02-0.18), pain (β = -0.12, CI = -0.21 to -0.06), and health-related quality of life (β = 0.11, CI = 0.04-0.21). There was an additional path to pain self-efficacy through pain-related coping (β = 0.06, CI = 0.01-0.12). Similar patterns were found in Models 2 and 3.
CONCLUSIONS: Improvements in stress and related constructs of depression and posttraumatic stress, and pain-related coping were causal mechanisms of effect in a physiotherapist-delivered integrated intervention. As integrated interventions are growing in popularity, it is important to further personalize interventions for improved benefit.

1. Introduction

Musculoskeletal pain conditions are among the leading causes of disability worldwide.[40] Motor vehicle crashes are a common cause of musculoskeletal pain, with neck pain or whiplash-associated disorders (WADs) being the most frequent injury. Up to 50% of people with whiplash injuries do not fully recover[4] and will experience a complicated recovery trajectory characterized by chronic pain, mental health problems, and disability.[16,35]

There are few effective treatments for acute WAD. Physical rehabilitation has only small effects[18,33,37] and early multidisciplinary treatment is no more effective than usual care.[15] In contrast, recent evidence indicates that identifying patients at risk of poor recovery and targeting both psychological and physical factors delivers more beneficial effects. The StressModex randomised clinical trial (RCT) compared a 6-week physiotherapist-delivered integrated psychological and exercise intervention to 6 weeks of exercise alone. The integrated intervention delivered greater decreases in pain-related disability (primary outcome) as well as improvements in secondary outcomes of pain intensity, pain self-efficacy, and quality of life that were clinically relevant and sustained for 12 months.

The psychological component of the integrated intervention aimed to target early stress based on evidence that acute stress is an important psychological risk factor for poor recovery.[3,17] Termed stress inoculation training, techniques including education about the effects of stress on pain and recovery, relaxation strategies, problem-solving and coping self-statements teach individuals to manage stress (see supplementary materials of the main trial,[36] http://links.lww.com/PR9/A71). It applies cognitive behavioral theory, which hypothesizes that treatments work through therapeutic changes in emotions (ie, stress) and cognitive and behavioral patterns (eg, coping; catastrophizing). Techniques were delivered within a self-management framework, enhancing individuals' confidence in managing stress and their recovery.

The approach of identifying “at-risk” patients and targeting early psychological risk factors in addition to physical rehabilitation is gaining momentum. In addition to WAD, studies of low back pain and injured workers have demonstrated clinically relevant benefits from this approach.[9,12,27] To date, no study has tested the theoretically relevant mechanisms of therapeutic change (mediators) in these integrated interventions. Information about change processes would help to inform refinement of theoretical models, clinical decision-making, design of more personalized treatment approaches,[2,19] and targeted skills training of care providers.

The aim of this study was to examine the potential causal mechanisms by which an integrated intervention for acute WAD improves health outcomes through a secondary mediation analysis of data from the StressModex RCT. Stress was a key intervention target and the primary mediator to be tested using the stress subscale of the Depression Anxiety and Stress Scale (DASS-21).[22] We hypothesized that the effects of the integrated psychological and exercise intervention on disability, pain-related self-efficacy, pain intensity, and health-related quality of life at 12 months would be mediated by reductions in levels of stress. We also considered potential cognitive behavioral mediators that had significantly improved 6-weeks posttreatment. It was also hypothesized that in addition to stress, pain-related coping would be a parallel mediator of the relationship between the integrated intervention and pain-related outcomes. In subsequent models, stress-related constructs that commonly co-occur with stress and pain and also share features of stress (eg, physiological arousal; anger/irritability) were tested as alternative mediators. In each model, parallel mediators were depressive symptoms and pain-related coping; and posttraumatic stress symptoms and pain-related coping.

2. Method

2.1. Participants and setting

The StressModex RCT was prospectively registered (ACTRN12614001036606) and the protocol[29] and results published.[36] The study received ethical approval from the human ethics research committees of The University of Queensland (2011000206), Griffith University (AHS/14/14/HREC), and the Gold Coast University Hospital (HREC/15/QGC/34). The sample has been previously described in detail.[36] Briefly, inclusion criteria were: acute (<4 weeks) WAD grade II/III (no fracture/dislocation), at least moderate neck pain-related disability (≥32%, on the Neck Disability Index[39]), and hyperarousal symptoms (≥3 on the hyperarousal subscale of the Posttraumatic Diagnostic Scale[7]). Participants were excluded if there was known or suspected spinal cord injury; confirmed fracture or dislocation; fracture or injuries to other body areas; spinal surgery in the past 12 months; met the criteria for probable acute stress disorder or major depression; or a history of psychosis, bipolar disorder, or depression. Refer to the methods of the main trial[36] for a summary of the number of individuals excluded due to severe levels of stress or major depression. After completion of baseline assessment measures, 108 patients were randomly allocated to receive the physiotherapist-delivered integrated stress inoculation and exercise intervention (n = 53) or exercise only (n = 55). The mean age of the sample was 41 years (SD = 14.1), the majority of participants were female (68%), had at least 10 years of schooling (77%), and were employed (64%). Outcomes were assessed at 6-weeks posttreatment, and at 6- and 12-month follow-ups. On the primary outcome, follow-up rates were 94% and 87% at 12 months for the integrated intervention vs physiotherapy exercise only group, respectively, with some variation in completion of secondary outcomes measures. Patient adherence to treatment was good; median number of sessions for both groups was 10 out of a possible 10 sessions. Physiotherapist adherence to the session protocols was excellent in both groups (>93%).[36]

2.2. Interventions

Detailed information on the interventions including the treatment manual can be found in the primary report and supplementary materials (http://links.lww.com/PR9/A71).[36] In short, the integrated intervention consisted of 6 physiotherapist-delivered sessions of stress inoculation and 10 sessions of physiotherapy exercise delivered over 6 weeks. The stress inoculation intervention comprised 3 main components: (1) identifying and understanding stress and its influence on pain and recovery, (2) developing skills for managing stress, and (3) applying skills in various situations to develop tolerance and confidence. Participants were encouraged to practice these skills each week during home practice. The physiotherapy exercise programme adhered to the Australian Guidelines for the Management of Acute Whiplash[34] and included exercises to improve movement, strength, and endurance of the neck and shoulder girdle muscles, eye/head coordination, and balance. Participants were encouraged to practice the exercises at home on a daily basis. The physiotherapy exercise only group received the 10-session exercise programme only.

2.3. Measures

For this secondary post hoc analysis, we chose all significant outcomes as in the main analyses except for the global impression of recovery scores because these were only collected immediately after treatment. These included 4 pain-related outcomes measures at 12 months: the Neck Disability Index,[39] Pain Self-Efficacy Questionnaire,[28] average pain intensity in the last 24 hours (0–10 scale), and the health-related quality of life (HRQoL) mental health component score from the SF-36.[10] The mediating variables at 6 weeks included the following: stress subscale from the DASS-21,[22] as well as the DASS-21 depression subscale,[22] the Posttraumatic Diagnostic Scale symptom score,[7] and the Coping Strategies Questionnaire.[30] These measures have been previously described in detail.[36]

2.4. Summary of previously reported StressModex randomised clinical trial results

As previously reported,[36] a linear mixed modelling and intention-to-treat analysis was used. Each outcome was analysed separately. Findings at 6-weeks posttreatment and 12-months follow-up are reported here because these are the time-points of interest for this secondary analysis. A significant and clinically important between-group difference in the primary outcome of pain-related disability was found, favoring the integrated intervention at 6 weeks and 12 months. Table 1 provides a summary. At 6 weeks, there was a significant treatment effect on secondary outcomes of posttraumatic stress symptoms, depression, stress, pain-related coping skills, pain intensity and HRQoL (mental health), but no effect on pain catastrophizing or anxiety symptoms. Significant treatment effects were also found at 12 months for pain intensity, pain self-efficacy, pain-related coping skills, and HRQoL (mental health).

Table 1

Treatment effects reported in the primary study at 6 weeks and 12 months.[35]

	6 weeks		12 months
	Mean difference	95% CI	Mean difference	95% CI
Primary outcome
NDI (0–100)	−10.0*	−15.5 to −4.48	−10.1*	−16.3 to −4.0
Secondary outcomes
PDS (0–51)	−3.7†	−6.9 to 0.5	−3.2	−6.7 to 0.4
DASS-stress (0–42)	−5.6‡	−8.4 to −2.8	−2.2	−5.1 to 0.8
DASS-anxiety (0–42)	−2.3	−4.6 to 0.1	−0.8	−3.7 to 2.1
DASS-depression (0–42)	−3.8*	−6.5 to −1.1	−2.5	−5.8 to 0.9
PCS (0–52)	−2.5	−5.6 to 0.5	−1.2	−4.8 to 2.4
PSEQ (0–60)	3.9	−1.2 to 9.0	6.7†	1.3 to 12.0
CSQ (−72 to 180)	12.9†	0.8 to 25	16.8*	4.8 to 28.8
HRQoL (MCS)	6.6*	2.1 to 11.2	4.5	0.2 to 8.7
24-hour pain intensity (0–10)	−1.5*	−2.3 to −0.6	−1.0†	−2.0 to −0.1

Baseline and 6-month measures omitted for clarity; results from linear mixed models with each outcome analysed separately, participant_id as the random effect and models adjusted for baseline values.

P < 0.001.

P < 0.01.

P < 0.05.

CI, confidence intervals; CSQ, Coping Strategies Questionnaire; DASS, Depression, Anxiety, and Stress Scale; HRQoL (MCS), Health-related Quality of Life (Mental Health Component scores) from the SF-36; NDI, Neck Disability Index; PDS, Posttraumatic Diagnostic Scale; PCS, Pain Catastrophizing Scale; PSEQ, Pain-related Self-Efficacy Questionnaire.

2.5. Analyses

Only variables with treatment effects at 6 weeks were tested as potential mediators. These were: stress symptoms (primary mediator) and pain-related coping skills as well as stress-related constructs of depressive and posttraumatic stress symptoms. Pearson correlations were performed in STATA (version 16) to examine potential multicollinearity among the mediators at 6 weeks (Table 2). Because stress, depressive, and posttraumatic stress symptoms at 6 weeks were highly correlated with each other (r values between 0.71–0.77; again, Table 2), it was not possible to enter them into a single model. We therefore created 3 separate simplex lagged parallel mediation models. In Model 1, stress symptoms and pain-related coping were entered as mediators; depressive symptoms and pain-related coping were entered in Model 2; and posttraumatic stress symptoms and pain-related coping were entered in Model 3. We used structural equation modelling to assess whether the primary mediator stress, or alternative mediators depressive or posttraumatic stress symptoms, and pain-related coping skills mediated the treatment effects (integrated intervention vs exercise control) on pain-related disability, pain self-efficacy, pain intensity, and HRQoL (mental health) at 12 months with the outcomes tested simultaneously. Structural equation modelling allows testing of multiple mediators and/or dependent variables simultaneously. For consistency with the primary analysis, we adjusted for baseline variables. The direct effect of the integrated intervention vs exercise control on each outcome (controlling for the mediators) and the average causal mediation effect (indirect effect) of the integrated intervention vs exercise control on 12-month health outcomes that is mediated by the hypothesized mediators were estimated. Indirect effects were computed using 95% confidence intervals (CIs) using bootstrapped standard errors (1,000 iterations), with estimates considered significant when CIs did not include zero. Standardized and unstandardized estimates, CIs, and R2 values were reported, as well as proportions of specific and total indirect effects. Absolute model fit was assessed and is represented by a nonsignificant χ2 statistic. Comparative fit indices were also assessed: comparative fit index (CFI), root mean-square error of approximation (RMSEA), and the standardized root mean-square residual (SRMR).[13] Cutoffs ≥0.95 for CFI, ≤0.06 for RMSEA, and ≤0.08 for SRMR represent good fit.[24] We note that these cutoffs are considered only as a guide, and models approaching these values were interpreted as having acceptable fit.[24] Missing data were handled using list-wise deletion. Mediational analyses were performed in MPlus Version 8.3.[26]

Table 2

Bivariate correlations of potential mediators at 6 weeks.

	1	2	3
1 Depression subscale (DASS)	1.00
2 Stress subscale (DASS)	0.77*	1.00
3 Posttraumatic stress (PDS)	0.71*	0.73*	1.00
4 Pain-related coping (CSQ)	−0.22†	−0.24†	−0.09

Total sample was used.

P < 0.001.

P < 0.05.

CSQ, coping strategies questionnaire; DASS, Depression, Anxiety, and Stress Scale; PDS, Posttraumatic Diagnostic Scale.

3. Results

After excluding cases due to missing data at follow-up, 103 of the 108 participants in the StressModex RCT were included. Standardized and unstandardized parameter estimates of the direct and indirect effects, significance levels, and CIs as well as proportions of the total effects are provided in Table 3 and Figures 1–3. Results are reported separately for each model.

Table 3

Direct, indirect, and total effects.

	Outcomes at 12 mo
	Pain-related disability			Pain-related self efficacy			Pain intensity			HRQoL (mental health)
	B	β	95% CI	B	β	95% CI	B	β	95% CI	B	β	95% CI
Model 1
Direct	−4.03	−0.11	−0.27 to 0.08	1.44	0.05	−0.14 to 0.22	−0.41	−0.08	−0.26 to 0.10	1.22	0.05	−0.14 to 0.23
Indirect through stress	−4.49	−0.12	−0.21 to −0.06	2.70	0.09	0.02 to 0.18	−0.61	−0.12	−0.21 to −0.06	2.64	0.11	0.04 to 0.21
Indirect through coping skills	−0.42	−0.01	−0.06 to 0.02	1.89	0.06	0.01 to 0.12	−0.14	−0.03	−0.08 to 0.01	0.23	0.01	−0.03 to 0.05
Total indirect	−4.92	−0.14	−0.24 to −0.06	4.58	0.14	0.05 to 0.25	−0.75	−0.15	−0.24 to −0.08	2.87	0.12	0.04 to 0.22
Total	−8.94	−0.25	−0.40 to −0.08	6.02	0.19	0.04 to 0.35	−1.16	−0.24	−0.41 to −0.06	4.09	0.17	0.01 to 0.33
Proportion mediated	55%			76%			64%			70%
Model 2
Direct	−6.26	−0.17	−0.34 to 0.01	2.78	0.09	−0.07 to 0.26	−0.73	−0.15	−0.34 to 0.03	2.06	0.08	−0.08 to 0.26
Indirect through depression	−3.24	−0.09	−0.16 to −0.02	1.76	0.06	0.01 to 0.12	−0.39	−0.08	−0.15 to −0.02	2.37	0.10	0.03 to 0.18
Indirect through coping skills	−0.27	−0.01	−0.06 to 0.03	1.84	0.06	0.01 to 0.12	−0.13	−0.03	−0.07 to 0.01	0.09	0.004	−0.04 to 0.04
Total indirect	−3.52	−0.10	−0.18 to −0.03	3.60	0.11	0.04 to 0.20	−0.52	−0.11	−0.18 to −0.04	2.46	0.10	0.02 to 0.18
Total	−9.77	−0.27	−0.43 to −0.10	6.39	0.20	0.05 to 0.36	−1.25	−0.25	−0.43 to −0.08	4.52	0.19	0.03 to 0.36
Proportion mediated	36%			56%			41%			54%
Model 3
Direct	−6.91	−0.19	−0.34 to −0.03	3.15	0.10	−0.06 to 0.26	−0.80	−0.16	−0.33 to 0.01	2.59	0.11	−0.05 to 0.27
Indirect through posttraumatic stress	−2.54	−0.07	−0.14 to −0.01	1.57	0.05	0.01 to 0.11	−0.38	−0.08	−0.15 to −0.02	1.92	0.08	0.02 to 0.15
Indirect through coping skills	−0.60	−0.02	−0.07 to 0.02	1.99	0.06	0.01 to 0.13	−0.16	−0.03	−0.08 to 0.002	0.31	0.01	−0.03 to 0.05
Total indirect	−3.14	−0.09	−0.17 to −0.03	3.55	0.11	0.04 to 0.20	−0.54	−0.11	−0.20 to −0.04	2.23	0.09	0.02 to 0.16
Total	−10.05	−0.27	−0.43 to −0.11	6.71	0.21	0.06 to 0.37	−1.34	−0.27	−0.44 to −0.09	4.81	0.20	0.04 to 0.36
Proportion mediated	31%			53%			40%			46%

Significant effects are shown in bold; unstandardized (B) and standardized (β) estimates shown. Bootstrapped 95% CIs for indirect paths.

Proportion mediated is the unstandardized total indirect effects divided by the total effects.

CI, confidence intervals; HRQoL, Health-related Quality of Life.

Figure 1.

Model 1: Paths between the integrated intervention and 12-month outcomes through changes in stress and coping skills at 6 weeks; solid lines represent significant paths; dotted lines are nonsignificant; HRQoL (MCS), health-related quality of life (mental component score). Baseline variables not shown for clarity.

Figure 3.

Model 3: Paths between the integrated intervention and 12-month outcomes through changes in posttraumatic stress and coping skills at 6 weeks; solid lines represent significant paths; dotted lines are nonsignificant; HRQoL (MCS), health-related quality of life (mental component score). Baseline variables not shown for clarity.

3.1. Model 1: testing stress and pain-related coping as mediators

The results of Model 1 are depicted in Figure 1. In line with the findings of the StressModex RCT, there was a direct intervention effect on 6-week stress and pain-related coping favouring the integrated intervention. There was a direct effect of stress on pain-related disability, pain self-efficacy, pain intensity, and HRQoL (mental health) at 12 months and a direct effect of pain-related coping on pain self-efficacy only. The integrated intervention had indirect effects on pain-related disability (β = −0.12, CI = −0.21 to −0.06), pain self-efficacy (β = 0.09, CI = 0.02–0.18), pain intensity (β = −0.12, CI = −0.21 to −0.06), and HRQoL (mental health) (β = 0.11, CI = 0.04–0.21) at 12 months through changes in levels of stress at 6 weeks. Changes in stress symptoms explained 50%, 45%, 53%, and 64% of the total integrated intervention effect on disability, pain-related self-efficacy, pain intensity, and quality of life, respectively. Indirect intervention effects on pain self-efficacy at 12 months through changes in pain-related coping skills at 6 weeks were also found (β = 0.06, CI = 0.01–0.12). Approximately 31% of the total effect of the integrated intervention on pain-related self-efficacy was mediated by coping. The direct effects of the integrated intervention on pain-related disability, pain self-efficacy, pain intensity, and HRQoL (mental health) at 12 months were not significant. Model 1 showed a good fit for the data (χ2 (31) = 47.30, P = 0.03, CFI = 0.96, RMSEA = 0.07, SRMR = 0.08). The total variance accounted for by each of the health outcomes ranged from 30% to 46% in Model 1. Refer to Table S1 in the supplementary materials for R2 values for the mediators and outcome variables (available at http://links.lww.com/PR9/A71).

3.2. Model 2: testing depression and pain-related coping as mediators

The results of Model 2 are depicted in Figure 2. There was a direct effect of the integrated intervention on depression and pain-related coping skills at 6 weeks. There were direct effects of depression on pain-related disability, pain self-efficacy, pain intensity, and HRQoL (mental health) at 12 months and a direct effect of pain-related coping on pain self-efficacy only. The integrated intervention showed indirect effects on pain-related disability (β = −0.09, CI = −0.16 to −0.02), pain self-efficacy (β = 0.06, CI = 0.01–0.12), pain intensity (β = −0.08, CI = −0.15 to −0.02), and HRQoL (mental health) (β = 0.10, CI = 0.03–0.18) at 12 months through changes in levels of depression at 6 weeks. Changes in depressive symptoms explained 33%, 27%, 31%, and 52% of the total integrated intervention effect on disability, pain-related self-efficacy, pain intensity, and quality of life, respectively. Indirect intervention effects on pain self-efficacy at 12 months were also found through changes in pain-related coping skills at 6 weeks (β = 0.06, CI = 0.01–0.12). Improvements in coping skills accounted for approximately 29% of the total integrated intervention effect on pain-related self-efficacy. There were no significant direct effects of the integrated intervention on disability, pain self-efficacy, pain intensity, and HRQoL. Model 2 showed a good fit for the data (χ2 (31) = 44.75, P = 0.05, CFI = 0.97, RMSEA = 0.07, SRMR = 0.08). The total variance accounted for by each of the health outcomes ranged from 27% to 45% in Model 2 (refer to Table S1 in the supplementary materials for R2 values, available at http://links.lww.com/PR9/A71).

Figure 2.

Model 2: Paths between the integrated intervention and 12-month outcomes through changes in depression and coping skills at 6 weeks; solid lines represent significant paths; dotted lines are nonsignificant; HRQoL (MCS), health-related quality of life (mental component score). Baseline variables not shown for clarity.

3.3. Model 3: testing posttraumatic stress and pain-related coping as mediators

The results of Model 3 are depicted in Figure 3. There was a direct effect of the integrated intervention on posttraumatic stress symptoms and pain-related coping skills at 6 weeks. There were direct effects of posttraumatic stress symptoms on pain-related disability, pain self-efficacy, pain intensity, and HRQoL (mental health) at 12 months and a direct effect of pain-related coping on pain self-efficacy only. There were indirect effects of the integrated intervention on pain-related disability (β = −0.07, CI = −0.14 to −0.01), pain self-efficacy (β = 0.05, CI = 0.01–0.11), pain intensity (β = −0.08, CI = −0.15 to −0.02), and HRQoL (mental health) (β = 0.08, CI = 0.02–0.15) at 12 months through changes in posttraumatic stress symptoms at 6 weeks. Changes in posttraumatic symptoms explained 25%, 23%, 28%, and 40% of the total integrated intervention effect on disability, pain-related self-efficacy, pain intensity, and quality of life, respectively. There was an indirect intervention effect on pain self-efficacy at 12 months through changes in pain-related coping skills at 6 weeks (β = 0.06, CI = 0.01–0.13). Approximately 30% of the total effect of the integrated intervention on pain-related self-efficacy was mediated by coping skills. The direct effect of the integrated intervention on pain-related disability at 12 months was significant but not on pain intensity, pain self-efficacy, and HRQoL. Model 3 showed a good fit for the data (χ2 (31) = 37.51, P = 0.20, CFI = 0.98, RMSEA = 0.05, SRMR = 0.07). The total variance accounted for by each of the health outcomes ranged from 33% to 48% (again, refer to Table S1 in the supplementary materials, available at http://links.lww.com/PR9/A71).

4. Discussion

Establishing effectiveness of integrated psychological and physical treatments for musculoskeletal pain in RCTs is important. Also important is to determine how these treatments work according to theory or other yet unidentified mechanisms. This secondary analysis of an RCT examined the underlying change mechanisms of an integrated physiotherapist-delivered psychological and exercise intervention for patients at risk of poor recovery after acute whiplash injury. Overall, the results support our hypothesis that improvements in multiple health outcomes are mediated by improvements in the intended intervention target of stress, as well as improvements in pain-related coping, which led to improvements in pain self-efficacy only. Stress-related constructs of depressive and posttraumatic stress symptoms were also important mechanisms of effect in an integrated intervention on health outcomes.

Intervention approaches that target early psychological risk factors known to delay recovery in addition to physical rehabilitation can lead to improved health outcomes for individuals with musculoskeletal pain.[9,12,27,36] No studies have been conducted examining if integrated psychological and exercise interventions for acute WAD exert their effects through intended intervention targets. Consistent with the StressModex RCT,[36] stress, depression, posttraumatic stress symptoms, and pain-related coping skills improved at 6 weeks after the integrated intervention compared with physiotherapy exercise alone. Stress was shown to be an important causal mechanism of therapeutic improvement, explaining 45% to 64% of the total effects of the integrated intervention on pain-related disability, pain intensity, pain self-efficacy, and HRQoL; 31% of the total intervention effects on pain self-efficacy were mediated by improvements in pain-related coping. The findings suggest that integrated interventions designed to target stress as a psychological risk factor after acute WAD can be beneficial for achieving a range of patient-relevant recovery outcomes. Components of the integrated intervention were explicitly designed to target levels of stress by enhancing patients' ability to cope with stress associated with pain and injury. For example, relaxation strategies were incorporated to address physiological arousal, and positive coping statements were used to address negative cognitions.

Broadly, the results of this study are in line with cognitive behavioral theory and previous investigations of psychological change processes over the course of cognitive behavioral therapies including acceptance and commitment therapy for chronic pain conditions.[21,38] Generally, in these studies, psychological interventions are delivered by psychologists. Our findings suggest that physiotherapists as well as psychologists can elicit changes in psychological variables in patients who have mild to moderate baseline scores and that these are the mechanisms by which the integrated intervention worked. They add to the emerging research in low back pain, which shows that the effects of psychologically informed vs usual primary care physiotherapy are associated with improved disability through changes in pain-related distress and pain severity.[23]

Causal mechanisms of change had consistent effects on health outcomes in each model. Similar to the pattern of findings in Model 1, 27% to 52% of the total effects of the integrated intervention on the 4 outcomes at 12 months were mediated by reductions in depressive symptoms and 29% of the total intervention effects on pain self-efficacy by improvements in pain-related coping (Model 2); 23% to 40% of the total integrated intervention effects on health outcomes were mediated by posttraumatic stress and 30% of the total integrated intervention effects on pain self-efficacy by pain-related coping (Model 3). These findings suggest that a number of stress-related constructs including depressive and posttraumatic stress symptoms are underlying mechanisms of therapeutic improvement for acute WAD. Measures of stress, depression, and posttraumatic stress symptoms used in this study were highly correlated, indicating a significant degree of overlap. Stress occurs when perceived environmental demands exceed the adaptive capacity to cope while measures of depression and posttraumatic stress capture more individual differences in affective and behavioral experiences suggestive of psychopathology.[5,6,20] Perceptions of stress are not the same construct as depressive or posttraumatic stress symptoms, although there is overlap.[6] Previous research has shown depressive symptoms and posttraumatic stress symptoms to be highly correlated.[8] However, validation studies of the DASS in clinical samples indicate correlations between 0.45 and 0.66 and distinguish between 3 separate factors.[1] Other research using community samples suggests there may be a common underlying factor of psychological distress as well as specific depression, anxiety, and stress factors.[11]

Changes in pain-related coping skills were also shown to mediate the effects of the integrated intervention but on pain self-efficacy only. For patients who received the integrated intervention, improvements in their ability to cope with pain led to improvements in their beliefs about their ability to perform certain activities when experiencing pain but not their functional ability to carry these out, or levels of pain. In the integrated intervention, coping strategies such as relaxation and developing coping statements aimed to help patients manage stress rather than pain. One session consisted of cognitive skill development, which involved patients developing positive coping statements to help them manage stressful situations. Coping strategies measured by the Coping Strategies Questionnaire[30] such as using distraction, ignoring sensations, and reinterpreting sensations were not included in the sessions. Therefore, it is possible that the integrated intervention failed to cause sufficient change in pain-related coping skills to improve disability-related outcomes. Previous research has suggested that changes in pain coping may not mediate specific treatment outcomes.[14,38] Failure to find mediation effects for pain-related coping on pain and disability does not mean it is not a critical mechanism. It is possible that coping with daily stressors and problems not measured in this study could be important. Temporal ordering of coping and stress-related variables or interactions between mediators could be considered in future studies. Finally, it may be possible to produce functional gains if pain-related coping skills are increased using other intervention techniques.

4.1. Limitations

This study had some limitations. We adjusted for baseline variables but did not control for possible confounders such as the number of treatment sessions patients completed or the extent to which physiotherapists adhered to the treatment protocol. As reported in the main trial, treatment completion and therapist adherence were high (>90%). The relatively small sample size may have resulted in insufficient power to detect small indirect effects through pain-related coping on other outcome variables. To enhance precision, we used bootstrapping methods to estimate 95% CIs, which is recommended when estimating indirect effects in mediation models with small to moderate sample sizes.[31] Because mediational analyses were designed post hoc, other potential cognitive behavioral and physical mediating factors were not measured. Due to multicollinearity, we did not include stress, depression, and posttraumatic stress mediators in a single model and were unable to account for the effects of correlated measurement error. We only investigated changes in mediating factors at 6-weeks postintervention and did not account for subsequent changes that may have occurred over time. The results are also limited to individuals with acute WAD who have mild to moderate symptoms of stress.

4.2. Strengths and future directions

This is one of the first studies to apply causal test of mediation to examine mechanisms of effect within the context of an RCT for acute WAD. It contributes to the emerging evidence base in favour of a theory-driven CBT approach by showing that changes in stress and pain-related coping are the processes by which a physiotherapist-delivered integrated psychological and exercise intervention worked. We assessed more than one mediator and included cognitive behavioral and alternative stress-based mediators. We also studied the effect of the mediators on multiple recovery outcomes and were able to show that improvements in self-efficacy can be reached through more than one path.

The results suggest that the integrated intervention effects could also be mediated through other pathways. Potential mediating variables such as pain-related fear known to be associated with poor recovery[25] and general (process) variables such as the clinician–patient alliance and patient expectations could be tested. Because the integrated intervention also included 10 exercise sessions, changes in measures targeted by this component of the intervention such as strength, coordination, and physical activity levels could be explored. Future research could consider incorporating objective measures of physiological stress reactivity as well as self-report measures of stress in childhood and cumulative life stress. Despite research showing that reductions in pain catastrophizing can mediate improvements in functioning,[32] it is possible that it might be a mediator in both integrated interventions and physical therapy only interventions because we did not find pain catastrophizing scores were different between the 2 groups. Possible shared vs specific mechanisms associated with the integrated and exercise interventions require further attention. Future RCTs should plan and design a priori mediational analyses to identify potential causal mechanisms. Potential moderators could also be tested such as number of completed sessions. Although the stress inoculation training did not specifically address depressive or posttraumatic stress symptoms, the findings also suggest that patients with acute WAD and at risk of poor recovery could benefit from a more targeted approach that specifically addresses these risk factors. This information could guide the development of more personalized integrated treatment approaches.

5. Conclusions

Improvements in stress, and related constructs of depressive and posttraumatic stress symptoms, as well as pain-related coping were causal mechanisms through which a physiotherapist-delivered integrated intervention exerted its effects on multiple health outcomes. Knowledge that reducing stress is an important treatment target in integrated interventions is useful for patients with WADs, clinicians, and researchers to understand how treatments work, or why they can fail. It also guides which skills of physiotherapists need to be trained. Given scarce health resources, we argue that shifting attention towards further developing and personalizing components of integrated interventions and designing future RCTs with a priori mediational analyses is critical to optimizing treatment effects for individuals with whiplash.

Disclosures

The authors have no conflicts of interest to declare.

Funding for the initial StressModex RCT was received from the National Health and Medical Research Council of Australia (Grant ID: APP1069443).

Supplemental digital content associated with this article can be found online at http://links.lww.com/PR9/A71.