Evidence of Physiotherapy Interventions for Patients with Chronic Neck Pain: A Systematic Review of Randomised Controlled Trials

Chronic neck pain (CNP) is common and costly, and the effect of physiotherapeutic interventions on the condition is unclear. We reviewed the literature for evidence of effect of physiotherapy interventions on patients with CNP. Five bibliographic databases (MEDLINE, EMBASE, CINAHL, Cochrane Library, and PEDro) were systematically searched. Randomised, placebo and active-treatment-controlled trials including physiotherapy interventions for adults with CNP were selected. Data were extracted primary outcome was pain. Risk of bias was appraised. Effect of an intervention was assessed, weighted to risk of bias. 42 trials reporting on randomised comparisons of various physiotherapy interventions and control conditions were eligible for inclusion involving 3919 patients with CNP. Out of these, 23 were unclear or at high risk of bias, and their results were considered moderate- or low-quality evidence. Nineteen were at low risk of bias, and here eight trials found effect on pain of a physiotherapy intervention. Only exercise therapy, focusing on strength and endurance training, and multimodal physiotherapy, cognitive-behavioural interventions, massage, manipulations, laser therapy, and to some extent also TNS appear to have an effect on CNP. However, sufficient evidence for application of a specific physiotherapy modality or aiming at a specific patient subgroup is not available.


Introduction
Musculoskeletal disorders are threatening quality of life by having the potential to restrict daily activities, cause absence from work, and result in a change or discontinuation in employment. These disorders are expensive for society and for patients and are responsible for the highest number of healthy years lost [1][2][3][4]. The prevalence of chronic neck pain varies. The 12-month prevalence of pain typically ranges between 30% and 50%; the 12-month prevalence of activitylimiting pain is 1.7% to 11.5% [5]. The annual incidence of neck pain associated with whiplash varies greatly. Although 50% of whiplash victims recover in three to six months, 30% to 40% have persisting mild to moderate pain and 10% to 20% retain more severe pain [6]. It is a multifaceted phenomenon with physical impairment, psychological distress, and social dysfunction, which calls for an evidence-based, cost-effective rehabilitation treatment [7][8][9][10][11].
According to a Dutch study, 44% of patients with chronic neck pain visited their general practitioner (GP) with the condition during a twelve-month period; 51% of these were referred to physiotherapy treatment [12]. Knowledge of the actual effect of physiotherapy is therefore important and is anticipated to be reflected in the awareness of evidence-based practice among physiotherapists.
The Cochrane Collaboration has provided systematic reviews on the effect of massage for mechanical neck disorders [13], patient education for neck pain [14], electrotherapy for neck pain [15], mechanical traction for neck pain with or without radiculopathy [16], and conservative treatment for whiplash [17]. The overall conclusion has been that the evidence for these treatments is low and that no definite statements on the efficacy and clinical usefulness of these treatments can be made. A further Cochrane Review on the effect of manipulation and mobilisation of neck pain found low quality evidence that cervical and thoracic manipulations may provide pain reduction [18]. An additional Cochrane Review on the effect of exercises for mechanical neck disorders concluded that the summarised evidence indicates that there is a role for exercises in the treatment of acute and chronic mechanical neck pain plus headache but that the relative benefit of each type of exercise needs extensive research [19].
However, none of these reviews have covered the majority of commonly used physiotherapy modalities in one in order to get an overview of the subject. Besides, the effect of specific physiotherapy treatments in specific subgroups of chronic pain patients is an important topic which has not yet been examined. Clinicians and policy makers need evidence from research to inform and guide clinical practice and policy. Patients and researchers also need such information to support shared decisions and to set priorities for future research.
The aim of this study was to review the literature systematically and discuss the quality of evidence of commonly used physiotherapy interventions (exercise, manual therapy, and electrotherapy) aimed at improving outcomes (on pain, function, and quality of life) important for patients with chronic neck pain [20]. Neck pain was defined as pain located in the anatomical region of the neck [21]. Pain was considered chronic if it had persisted for more than three months, as defined by the International Association of the Study of Pain.

Methods
We performed a systematic review of all available randomised controlled trials on the subject of physiotherapy for neck pain to determine the effects of physiotherapy interventions on pain, function, and quality of life in neck-pain patients and to explore whether beneficial effects could be explained by biases affecting individual trials [22]. Study selection, assessment of eligibility criteria, and data extraction were carried out based on a predefined, peer-reviewed protocol according to the Cochrane Collaboration's guidelines [23]. This paper was prepared in accordance with the PRISMA statement [24].

Literature Search.
We searched five bibliographic databases (MEDLINE, EMBASE, CINAHL, Cochrane Library, and PEDro) from January 1990 to January 2012 with a structured, pre-defined, search strategy [25]. The search strategy was "Neck Pain AND Physiotherapy Intervention. " For neck pain, the following terms were combined with OR: "whiplash/WAD, " "neck injury, " "neck sprain/strain, " "neck ache, " "cervical sprain/strain, " "cervical disorder/syndrome, " "cervical spondylosis/itis, " "cervical osteoarthritis", "cervicodynia", "cervicobrachial pain/disorder/syndrome", "myofascial pain/disorder/syndrome, " "trapezius myalgia, " "postural syndrome, " and "nonspecific neck pain. " For physiotherapy interventions, the following terms were combined with OR: "physiotherapy, " "physical therapy, " "rehabilitation, " "intervention studies, " "exercise, " "exercise therapy, " "exercise movement techniques, " "manual therapy, " "manipulative medicine, " "mobilisation/mobilization, " "musculoskeletal techniques, " and "electric/electro stimulation therapy. " All terms were searched as free text as well as keywords, where this was applicable. Limitations were human studies in the English, German, Dutch, Danish, Norwegian, and Swedish languages, in the time span of January 1990 to January 2012. To assure that the included studies followed scientifically sound methods and the data therefore were well documented, we set a limit for inclusion to publications from 1990 and onwards.
Reference lists of review articles and included studies were searched to identify other potentially eligible studies. An additional search was conducted via the scientific search machine http://www.scirus.com/, using the following search terms combined with AND: "chronic neck pain, " "physiotherapy. "

Selection Criteria.
Studies were included if participants were older than 18 years of age and had chronic neck pain for more than three months (therefore considered chronic). Chronic neck pain was defined as (i) chronic whiplashassociated disorders (WAD); (ii) chronic non-specific neck pain, including work-related neck pain, myofascial neck pain, upper trapezius myalgia, chronic neck pain associated with degenerative findings with or without radicular findings, or other surrogate terms.
Eligible interventions were physiotherapy interventions commonly used in the treatment of musculoskeletal pain: (i) exercise therapy, including specific types of exercises, for example, neuromuscular training, strength training, and endurance training; (ii) manual therapy, for example, massage, manipulations, and mobilisations; (iii) electrotherapy, for example, TENS, low-level laser, or other surrogate terms. Acupuncture was not considered a physiotherapy technique since this technique is not part of physiotherapy in all countries. Comparison of the therapy had to be made with no treatment (e.g., waiting list controls), or other conservative active therapies called "care as usual, " or sham therapy. Anticipating that only a limited number of trials available used placebo/sham control, we decided also to include trials in which an active control was used as a cointervention.
To be eligible for inclusion, a study must apply at least one pain measurement prior to and following the intervention, which was an outcome considered to be of major importance to the patients. Self-reported function and disability [26], self-reported quality of life [27], objective physical function, and clinical tests were considered minor outcomes and therefore not considered necessary inclusion criteria [28][29][30]. Only randomised controlled trials were accepted. Exclusion criteria were studies with participants with acute or subacute  neck pain, neck pain with definite or possible long tract signs, neck pain due to specific pathological conditions (e.g., fractures, tumours, infections, inflammatory processes, ankylosing spondylitis, and rheumatoid arthritis), and headache. We created a reliable process through consequently two reviewers who independently conducted the study selection and assessment of eligibility criteria. Similarly, two reviewers independently conducted data abstraction and assessed the risk of bias. Disagreements were resolved through consensus with a third reviewer being consulted if there was disagreement.

Data Extraction and Evidence
Synthesis. Data regarding publication status, trial design, patient characteristics, treatment regimens, outcome methods, results, and funding were extracted on a standardised form using a custom-made Microsoft Excel spreadsheet.
We assessed the risk of bias by using the Cochrane Collaboration's tool for assessing risk of bias as presented in [23]. Each of the following domains would be considered adequate-that is, presumably with a low risk of bias (i) "adequate sequence generation"; (ii) "allocation concealment"; (iii) "blinding"; (iv) "incomplete outcome data addressed"; (v) "free of selective outcome reporting"; (vi) "free of other bias (i.e., whether a study sponsor would benefit economically from a positive outcome). Each of these key components of methodological quality was assessed on an Adequate/Unclear/Inadequate basis. We used The Cochrane Collaboration's approach for summary assessments of the risk of bias for each important outcome across domains within a trial [23].
Due to the limited number of studies investigating each of the specific interventions, it was decided that both metaanalytical and level of evidence approaches would be inappropriate. Therefore, a narrative approach where we evaluated the study and results between groups within a trial was used to summarise the findings. To formulate conclusions, only results from trials at low risk were considered as evidence for an intervention.

Results
The literature search identified 4921 relevant studies (1110 from EMBASE, 1568 from MEDLINE, 1239 from CINAHL, and 491 from PEDro), of which 3685 were duplicates, leaving 1236 potentially eligible studies to be screened (see Figure 1). Following screening of titles and abstracts, 151 potentially relevant studies were identified and retrieved in full text. Finally, 42 randomised controlled trials, involving 3919 patients, fulfilled the selection criteria and were considered suitable for inclusion. The selection process and reasons for exclusions are presented in Figure 1.

Study
Characteristics. Study characteristics and study results are presented under the categories exercise therapy (25 trials, 18 regarding chronic non-specific neck pain, and seven regarding chronic neck pain related to whiplash); manual therapy (six trials, all related to chronic non-specific neck pain); and electrotherapy (11 trials, all related to chronic nonspecific neck pain) in Appendix A, Tables 1-4.
The trials covered the following intervention topics: (i) exercise therapy: various types of dynamic and isometric exercises, general aerobic exercises, exercises with a focus on strength, endurance, proprioception and coordination, specific neck stabilising exercises, craniocervical-flexion exercises, posture, behavioural graded activity, relaxation, body awareness, myo-feedback training, and multimodal physiotherapy; (ii) manual therapy: massage, manipulation, and traction; (iii) electrotherapy: laser, transcutaneous nerve stimulation (TENS), ultrasound, and repetitive magnetic stimulation (rMS).
Sham therapy or waiting list controls were used as control groups in 12 trials; 10 trials used a control group consisting of a self-management book, health-counselling, or other interventions, clearly distinguished from the active intervention group; six trials used active-treatment control reported as "treatment as usual"; active-treatment control was used in 14 trials.
Primary outcome measures were self-reported pain and/or self-reported pain and disability in 41 trials; when primary outcome measures were not reported, all outcome measures were considered. One trial had an objective test as primary outcome, yet pain was included in the secondary outcome measures.

Risk of Bias.
Risk of bias is presented in Appendix B, Table 5.
Overall, the quality of reporting on methodological issues varied. Table 5 shows the judgements ("Adequate, " "Unclear, " and "Inadequate") for each of the domains. As can be seen, 28 of 42 trials succeeded in reporting on adequate sequence generation; 18 trials described adequate allocation concealment; four trials adequately reported on attempts to blind participants, personnel, and outcome assessors; 22 trials adequately reported on missing outcome data, using intention-to-treat analysis; three trials adequately reported on selective outcome reporting by referring to a published and available protocol for comparisons; and 25 trials adequately reported on funding and the role of funding.
All studies are described in detail in Appendix A, Tables 1-4. All trials at low risk of bias, showing an effect on pain, are, furthermore, presented in the following section. According to the described criteria, the evidence for each intervention will following be summarised at the end of each section.

Exercise
Effect of Exercise on Pain in Patients with Chronic Nonspecific Neck Pain. As shown in Appendix A, Table 1, 18 trials examined the effect of various types of exercise in patients with chronic neck pain; nine of these were at unclear or high risk of bias [31-33, 40-44, 46], and nine were at low risk of bias [34-39, 45, 47, 48]. Seven of the trials at low risk examined the effect of different types of exercise, including proprioception exercises (eye-head coordination), craniocervical flexion exercises (C-CF), neck stabilisation exercises, stretching, strengthening, and behavioural graded activity programme, but did not find statistically significant difference on pain between groups following intervention [34, 36-39, 45, 48]. Two of the trials at low risk of bias succeeded in finding an effect on pain from the intervention, and for this reason, their results were considered evidence of use of exercise.
(1) Gustavsson et al. [35] examined a multicomponent pain and stress self-management group intervention (PASS) versus a control group receiving individually administered physiotherapy (IAPT). There was a statistically significant effect on ability to control pain ( < 0.001) and on neck-related disability (NDI) ( < 0.001) in favour of PASS at the 20-week followup.
(2) Ylinen et al. [47] examined three interventions: intensive isometric strength training versus lighter endurance training versus a control group. The two training groups had an additional 12-day institutional rehabilitation programme. At the 12-month followup, both neck pain and disability had decreased in both training groups compared with the control group ( < 0.01).
No trials with low risk of bias supported single use of proprioception exercises (eye-head co-ordination), craniocervical flexion exercises (C-CF), or neck stabilisation exercises for pain. No trials with low risk of bias support the use of stretching.
Effect of Exercise on Pain in Patients with Chronic Whiplash-Associated Disorder. As shown in Appendix A, Table 2, seven trials examined the effect of various types of exercise in patients with chronic WAD; three of these were at low risk of bias [49,51,55], and four were at unclear or high risk of bias [50,[52][53][54]. One of the trials at low risk of bias examined the effect of adding biofeedback training to a rehabilitation programme, but found no difference in effect between groups [49]. Two trials at low risk of bias succeeded in finding an effect on pain from the intervention, and for this reason, their results were considered evidence of use of exercise.
(1) Jull et al. [51] examined a multimodal physiotherapy programme (including exercises, education, and ergonomics) versus a self-management programme. The multimodal physiotherapy programme group attained a statistically significant greater reduction in reported neck pain and disability (NDI) posttreatment ( = 0.04). (2) Stewart et al. [55] examined exercise (e.g., endurance, strength, aerobic, coordination, and cognitive behavioural therapy) versus advice alone. Exercise and advice were more effective than advice alone at  Griffiths et al. [34] Chronic neck pain (diagnosed spondylosis, whiplash, nonspecific neck pain, and discogenic pain), age 18 and over ( = 74) (1) Specific neck stabilisation exercises ( = 37) in addition to the same programme as group 2 (2) General neck exercise programme ( = 37), posture correction technique, and active range of movement exercise All: max. four 30 min treatment sessions within the first 6 wks, advice to perform exercises 5-10 times daily, written sheets, after 6 wks the therapist could discharge the patient or continue NPDS, NPQ, VAS * * There were no significant between-group differences in the NPDS at either 6 wks or 6 mths 6 ISRN Pain    (2) Continued chiropractic treatments and were advised to do "phasic neck exercises" (eye-head co-ordination) ( = 15) All: exercises 5 days a wk for 8 wks NPDI The authors do not report any data on statistically significant differences between groups after 8 wk Jull et al. [51] Patients with chronic whiplash-associated disorder (>3 mths, <2 yrs), classified WADII,     VAS, pupil diameter No statistically significant differences between groups immediately after the treatment       No trials at low risk of bias support the use of EMG biofeedback.

Manual Therapy
Effect of Manual Therapy on Pain in Patients with Chronic Nonspecific Neck Pain. As shown in Appendix A, Table 3, six trials examined the effect of various types of manual therapy in patients with chronic non-specific neck pain [56][57][58][59][60][61]. One of the trials was at unclear risk of bias, and for that reason not considered evidence [61]. Three trials at low risk of bias examining the effect of spinal manipulations found no difference between groups [56,58,60]. Two trials succeeded in finding an effect on pain from the intervention. Both trials had a low risk of bias, and for this reason, their results were considered evidence of use of manual therapy.
(2) Sherman et al. [59] examined massage versus a selfcare book. They found statistically significant effect on massage following four weeks of treatment on neck pain and disability (NDI) ( = 0.047), but not at longterm followup at 10 and 26 weeks.
No trials at low risk of bias support the use of traction.

Electrotherapy
Effect of Electrotherapy on Pain in Patients with Chronic Nonspecific Neck Pain. As shown in Appendix A, Table 4, 11 trials examined the effect of various types of electrotherapy in patients with chronic non-specific neck pain; two of these were at low risk of bias [63,64], and nine were at unclear or high risk of bias [62,[65][66][67][68][69][70][71][72]. The two trials at low risk of bias both succeeded in demonstrating an effect on pain from this type of intervention and for this reason; their results were considered evidence of use of electrotherapy.
(1) Chiu et al. [63] examined three interventions: TENS versus exercise versus a control group. There were no statistically significant differences between the three groups on pain (VNPS) after 6-week and at 6-month followup, but the TENS group and the exercise group had a significantly better improvement in neck pain and disability (NPQ) than the control group ( = 0.034 and = 0.02, resp.) after 6-week, and at 6month followup.
(2) Chow et al. [64] examined laser versus sham laser treatment. The improvement in VAS was statistically significantly greater in the laser treatment group than in the sham laser treatment group (−2.7 compared with +0.3, < 0.001) at 12-week followup.
No trials at low risk of bias support the use of ultrasound therapy. No trials at low risk of bias support the use of rMS.

General Interpretation.
In this review, we assessed the effect of various interventions for the treatment of chronic neck pain and evaluated the methodological quality of the trials. Our findings emphasise the importance of taking the risk of bias into consideration when evaluating the evidence of an intervention.
Trials varied substantially regarding their internal validity, although the methodological quality of the RCTs in general appeared to be somewhat low with an unclear or high risk of bias. We identified various methodological flaws that may have implications for the internal validity of the trials and consequently may result in biased outcomes. Key domains in this context were randomisation, blinding, and incomplete outcome data.
Our evaluation also exposes a widespread use of withingroup analyses, claiming statistically nonsignificant results to be beneficial. Results were frequently analysed and reported as if they were uncontrolled within-group studies, which consequently led to misinterpretation of results. To some extent this may be due to the absence of a control group in many trials, and the use of an active treatment as a comparative group makes the "proof " of a truly statistically significant effect more difficult to find. We believe that attention should be paid to inadequate interpretation of a trial result when authors inadequately interpret lack of difference in terms of efficacy [73][74][75][76].

Effect of Physiotherapy on Chronic Neck Pain.
Overall, the evidence of effect of physiotherapy for chronic neck pain is strengthened. Yet, for some of the treatments offered, no definite effect and clinical usefulness can be shown. This does not necessarily implicate that these treatments have no effect, only that the present evidence is not sufficient.
Physiotherapy interventions for chronic neck pain showing the strongest support for an effect on pain are strength and endurance training (supported by two trials by Stewart et al. [55] and by Ylinen et al. [47], treating patients with chronic WAD and patients with chronic non-specific neck pain, resp.). In patients with chronic WAD, multimodal physiotherapy was also shown to have a beneficial effect by one trial by Jull et al. [51]. In patients with chronic non-specific neck pain, the use of cognitive/behavioural components in exercise was supported by one trial by Gustavsson et al. [35]. In regard to manual therapy, massage seems to have an effect on pain in patients with chronic non-specific neck pain, supported by one trial by Sherman et al. [59], and thoracic manipulation seems to have an effect on pain, supported by one trial by Lau et al. [57]. Within the area of electrotherapy, both laser therapy and TNS seem to have an effect on pain in patients with chronic non-specific neck pain. This was supported by one trial by Chow et al. [64] and one trial by Chiu et al. [63]. No trials supported the isolated use of proprioception (eye-head co-ordination), cranio-cervical flexion training, stretching, ultrasound therapy, rMS, and traction.
When looking deeper into the actual components of the various interventions in the above-mentioned trials, four of them-despite the differences, diversity, and individual features of the interventions-seem to have several characteristics in common: The interventions can be considered to be rehabilitative interventions of multimodal physiotherapy with a focus on exercise, including cognitive-behavioural components. This is based on (1) the trial by Stewart et al. [55] showing effect of mixed exercises, where the intervention besides submaximal training, stretching, and aerobic endurance included coordination programme designed to improve functional activities and principles of cognitive behavioural therapy (i.e., setting goals); (2) the trial by Ylinen et al. [47] showing effect of strength training and endurance training, where training groups had an additional 12-day institutional rehabilitation programme with training lessons, behavioural support, ergonomics, sessions of physical manual therapy-including massage/mobilisations-and advice to continue exercise; (3) the trial by Jull et al. [51] showing effect of a multimodal physiotherapy, including low-load exercise for reeducating muscle control of the neck flexor and extensor muscles and scapular muscles, posture exercises, kinaesthetic exercises, and mobilisation techniques, and in addition education including ergonomics, daily living advice, and home exercise; (4) finally, the trial by Gustavsson et al. [35] who found effect from a multi-component pain and stress self-management group intervention-including relaxation training, body awareness exercises, and lectures and group discussions-regarding coping with pain in terms of patients' self-reported pain control, self-efficacy, and disability. Our main results are consistent with findings of previous reviews of interventions for neck pain. The Cochrane Review by Kay et al. [19] on the effect of exercises for mechanical neck disorders concluded that the summarised evidence indicates that there is a role for exercises in the treatment of acute and chronic mechanical neck pain plus headache, but that the relative benefit of each type of exercise needs extensive research. Our review on chronic neck pain agrees with the present conclusion regarding exercise, yet our findings tend to favour strength and endurance training, as well as multimodal physiotherapy in addition to pain and stress selfmanagement. The superior effect of strength training and endurance training may be due to the physical impairments found in the chronic condition [77][78][79][80].
Our review adds new knowledge regarding the evidence for use of massage. Our findings are in discrepancy to a Cochrane Review by Haraldsson et al. [13] who concluded that the effectiveness of massage for improving neck pain and function remains. Yet the quoted review was last updated in 2004, and the trial by Sherman et al. [59] supporting massage was published in 2009. Our findings on the evidence of manipulation are in line with another Cochrane Review by Gross et al. [18] on the effect of manipulation and mobilisation for neck pain, who found low quality evidence that cervical and thoracic manipulations may provide pain reduction.
We too found evidence that thoracic manipulations may have an effect on pain [57]. Regarding low-level laser therapy, our findings are consistent with the findings of a review by Chow et al. [81], who found that low-level laser therapy reduces pain in patients with chronic neck pain. A Cochrane Review from 2007 [17] on the effect of conservative treatment for whiplash concluded that clearly effective treatments are not found for treatment of acute, subacute, or chronic symptoms. The findings of our newer review do support multimodal physiotherapy and mixed exercise programmes for chronic WAD. The explanation for the difference may be that the Cochrane Review by Verhagen in 2007 was not updated after January 2007, and our findings are based on more recent research, namely, two trials published later in 2007 [51,55]. A more recent review by Teasell et al. [82] found evidence to suggest that exercise programmes are the most effective noninvasive treatments for patients with chronic WAD. Our findings give support to the use of cognitive-behavioural element, and to pain and stress self-management. This is in discrepancy to another Cochrane Review by Gross et al. [14] on patient education for neck pain, concluding that there is no strong evidence for the effectiveness of educational interventions in various neck disorders. This difference may be due to the use of only single-modal trials in their review rather than multimodal trials as used in the current review.

Strengths and Weaknesses of Review Procedures.
To our knowledge, this is the first systematic review on interventions for chronic neck pain addressing the majority of commonly used physiotherapeutic modalities in one study, in order to get an overview of the subject area.
The search strategy and selection criteria we used were quite strict and easy to apply and according to normal procedures for conducting systematic reviews [23]. Yet the following limitations of the literature search may have introduced a bias: some relevant trials may have been missed if they used other keywords, although this is not very likely. We had limitations in language, and this may have led to missing studies from countries in Eastern Europe with a tradition of physiotherapy research, like Poland. We decided to limit our search from 1990 to January 2012. This was due to physiotherapists prior to this time not being trained in scientific methods and also that RCTs were rare. Studies earlier than 1990 would in general not be following a strict protocol like the ones used for RCTs, but at best be longitudinal cohort studies.
The quality assessment was presented in a reproducible manner. However, the results may be affected by our emphases during filtering methods for synthesis evidence. We might have chosen to exclude all trials with insufficient reporting on allocation sequence and allocation concealment. However, we chose not to, since this would have left us with very few trials to assess. We assessed risk of bias, requiring a convincing mechanism to be described in order for a trial to be classified as "adequate. " Our approach to this problem was to assume that the quality was inadequate unless information on the contrary was provided, and in doing so, we might have misclassified well-conducted but badly reported trials.
The present review succeeded in a subgroup assessment of physiotherapy treatment for chronic non-specific neck pain and for chronic WAD. Yet the first group was very wide due to the mixed conditions in the group of participants. The various interventions were considered to be complex, multifaceted, and with various cointerventions, and by classifying them into intervention groups according to-what we believed to be-the trial's agenda, we may have misclassified some. On the other hand, the often used combined therapies also highlight a fundamental problem when assessing effect of specific and single physiotherapy modalities. Another issue is the quality of the intervention since the interventions were administered in different ways and in different settings. It is reasonable to expect that the way in which they were administered including the dose-response relationship could have influenced the outcome. It would have been interesting and very relevant to examine this. Herbert and Bo [83] emphasise that researchers carrying out systematic reviews should routinely examine the quality of interventions.

Future Directions.
We need to know which patients will benefit from which intervention, built on well-conducted and well-reported trials, considering subgroups of patients with chronic neck pain, in order to support recommended evidence-based decisions and to set priorities for future research. We also request future trial investigators to consider to what extent cointerventions are valuable, in addition to possible confounders. Another issue to consider is the extent to which the control groups ought to be given care and attention to the same extent as the intervention groups.

A. Study Characteristics and Study Results
See Tables 1-4.

B. Risk of Bias
See Table 5.