Conservative management of mechanical neck pain: systematic overview and meta-analysis

BMJ 1996; 313 doi: (Published 23 November 1996) Cite this as: BMJ 1996;313:1291
  1. Peter D Aker, associate professora,
  2. Anita R Gross, assistant professorb,
  3. Charles H Goldsmith, professorc,
  4. Paul Peloso, assistant professord
  1. a Division of Graduate Studies and Research, Canadian Memorial Chiropractic College, Toronto, Ontario, Canada
  2. b Chedoke-McMaster Hospitals and School of Rehabilitation Sciences, McMaster University, Hamilton, Ontario, Canada
  3. c Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario
  4. d Department of Medicine, University of Saskatchewan, Saskatoon, Saskatchewan, Canada
  1. Correspondence to: Dr C H Goldsmith, Centre for Evaluation of Medicines, St Joseph's Hospital, 50 Charlton Avenue East, Hamilton, Ontario L8N 4A6, Canada.
  • Accepted 19 September 1996


Objective: To review the efficacy of conservative management of mechanical neck disorders.

Methods: Published and unpublished reports were identified through computerised and manual searches of bibliographical databases, reference lists from primary articles, and letters to authors, agencies, foundations, and content experts. Selection criteria were applied to blinded articles, and selected articles were scored for methodological quality. Effect sizes were calculated from raw pain scores and combined by using meta-analytic techniques when appropriate.

Results: Twenty four randomised clinical trials met the selection criteria and were categorised by type of intervention: nine used manual treatments; 12 physical medicine methods; four drug treatment; and three education of patients (four trials investigated more than one form of intervention). The intervention strategies were summarised separately. Pooling of studies was considered only within each category. Five of the nine trials that used manual treatment in combination with other treatments were combined. One to four weeks after treatment the pooled effect size was −0.6 (95% confidence interval −0.9 to −0.4), equivalent to an improvement of 16 (6.9 to 23.1) points on a 100 point scale. Sensitivity analyses on study quality, chronicity, and data imputation did not alter this estimate. For other interventions, studies could not be combined to arrive at pooled estimates of effect.

Conclusions: There is little information available from clinical trials to support many of the treatments for mechanical neck pain. In general, conservative interventions have not been studied in enough detail to assess efficacy or effectiveness adequately.

Key messages

  • Many treatments are available and accepted as standard forms of practice

  • Systematic literature searching finds a limited number of clinical trials

  • There is early evidence to support the use of manual treatments in combination with other treatments for short term pain relief, but in general, conservative interventions have not been studied in enough detail to assess efficacy or effec- tiveness adequately

  • Further clinical trials are needed to determine optimal treatment approaches


Neck pain is a common complaint, with a point prevalence of nearly 13%1 2 and lifetime prevalence of nearly 50%.3 4 5 In some industries neck related disorders account for as many days of absenteeism as low back pain.6 Many treatments are available to patients and accepted as standard forms of practice, including such common conservative strategies as medication, physical medicine methods, manual treatments, and education of patients.7 8 There is little evidence, however, for their accepted use. To determine the efficacy of the various conservative treatment strategies for mechanical neck pain our multidisciplinary team undertook this systematic overview. This article summarises a more detailed report published on line.9



Computerised bibliographical databases (MEDLARS, EMBASE, CINAHL, and CHIROLARS) and indexes of conference proceedings (NTIS and CPI) were searched without language restrictions from 1985 to December 1993. To reduce the likelihood of missing information published before this time, references lists of all retrieved articles were reviewed independently by two investigators to identify additional relevant citations. Letters were sent to first authors, relevant agencies and foundations, drug companies, and content experts to help to identify further published, unpublished, or current research.


The titles of citation printouts were reviewed by two investigators independently. All citations identified by either investigator as being potentially related to our topic were retrieved. The methods sections from these articles were reviewed by two investigators who had no access to the abstract, results, discussion, and conclusion sections. Authors of articles published in abstract form alone were contacted for full manuscripts and datasets. Randomised controlled trials included adults with mechanical neck disorders treated with conservative treatments and reported pain outcomes. For the purposes of this review conservative treatments were defined as any non-invasive, non-surgical form of treatment including drugs, manual treatments, education of patients, and physical medicine methods.8 Mechanical neck disorders included conditions causing neck pain with or without referral into the shoulder and upper arm.10 Studies investigating neck disorders with neurological deficit, headache without associated neck pain, or neck pain caused by other pathologies such as inflammatory disease, neurological disease, fracture, dislocation, neoplasm, or infection were excluded.


Three investigators independently assessed each selected study for methodological quality in five key validity criteria and assigned an overall score from 1 to 5. Lower scores (1 and 2) were assigned to weaker studies and higher scores (4 and 5) to stronger studies. The criteria and their scoring are outlined in detail elsewhere.9


Raw data on pain scores were extracted from the full manuscripts and were transformed to a 100 point scale. The effect size for each study was calculated as the mean difference between treatment and control groups divided by the pooled standard deviation. A fixed effects model11 was used to pool effect sizes with Metanal software.12 In studies where data were not available in a form to calculate an effect size the inverse 2 method (Fisher's method) was used to combine P values.11 When data could not be extracted and were not available from authors, values were imputed if possible.9

Studies were assigned into treatment categories and assessed for similarity before calculation of a pooled effect measure. To be considered eligible for pooling, trials must have studied subjects with similar disorders treated with similar interventions. If combining was sensible homogeneity of the estimates was formally tested with the Breslow-Day test.13 When data were not heterogenous a pooled effect size (with 95% confidence intervals) was estimated and the magnitude assessed.



The computerised search strategies located 917 citations. Sixty eight responses were received from over 200 letters sent out. In total, 200 citations were identified as being potentially related to mechanical neck disorders. Of these, 169 were excluded during the selection process: 127 did not study adults with mechanical neck disorders, 27 were not randomised controlled trials, 25 did not use conservative management strategies, and 20 did not report on pain outcomes (some studies were excluded for more than one reason). Twenty four randomised controlled trials identified from 31 publications met our selection criteria.14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44

DATA SYNTHESIS Manual treatments

Nine randomised controlled trials that used various types of manual treatments (including manipulation, mobilisation, and massage) met our eligibility criteria.17 18 23 25 33 35 36 39 43 Validity scores varied from strong to weak; criterion scores are reported in the summary tables.

Manual treatment alone—Two trials studied the use of manual treatments alone (that is, not in combination with any other form of treatment) compared with other treatments (table 1).18 43 Both compared manipulation with mobilisation, one rated as methodologically strong18 and the other moderate.43 Cassidy et al showed no significant difference between treatments.18 Vernon et al reported a significant improvement in the manipulation group.43

Table 1

Trials that used manual treatment for neck pain

View this table:

Manual treatment in combination with other treatments—Seven studies compared manual treatment in combination with other forms of treatment with another intervention (table 2.17 23 25 33 35 36 39 Nordemar et al36 and Sloop et al39 were included in this section because of potential interaction effects between the manual treatments and control treatments. In all seven studies manual treatments in combination with other treatments were compared with a control and considered eligible to pool. Brodin17 and Sloop et al39 were excluded from the pooled analysis because we were unable to obtain data to calculate effect sizes; the remaining five studies were combined. Control groups were relatively similar, consisting of rest, education, and use of analgesics with or without the use of exercise, cold packs, or cervical collars. The case mix included acute whiplash,33 35 acute36 or subacute and chronic mechanical disorders,25 and chronic cervical headache.23 All conform to the definition of a mechanical neck disorder and hence were considered to be clinically similar. As the duration of complaint (from acute to chronic) and the methodological quality varied between studies, sensitivity analysis was conducted. All outcomes were compared at similar end points—between one and four weeks of treatment. Specifically, the comparisons made were between the effect size for pain scores in Koes et al25 at three weeks of treatment, Mealy et al35 at four weeks of treatment, Nordemar et al36 at one week of treatment, Jensen et al23 at one week of treatment, and McKinney et al33 at four weeks of treatment. There was no evidence of heterogeneity between studies (Breslow-Day test with 4 df, P = 0.976), and the studies were combined to yield a pooled effect size of −0.6 (95% confidence interval −0.9 to −0.4) (fig 1). This pooled effect size is equivalent to a change of 16.2 points on a 100 point scale (6.9 to 23.1), which does not alter significantly when the study of lower methodological quality33 is dropped (−0.7; −1.0 to −0.3) or when those studies for which standard deviation was imputed25 33 were withdrawn (−0.5; −1.0 to 0.0). Similar point estimates of effect size were found for those trials that assessed acute (−0.63; −0.93 to 0.32) and subacute or chronic (−0.45; −1.05 to 0.15) patient groups. The small number of studies used in these subgroup analyses would be unlikely to have sufficient power for meaningful results.

Table 2

Trials of manual treatment combined with drug treatment, education, or physical medicine methods for treatment of neck pain

View this table:

Physical medicine methods

Twelve randomised controlled trials that used physical medicine methods (including spray and stretch, laser, electromagnetic treatment, infrared, acupuncture, traction, exercise, and transcutaneous electrical nerve stimulation (TENS)) met our eligibility criteria.20 21 22 30 31 32 36 37 38 40 41 42 43 44 Validity scores varied from strong to weak. Table 3 summarises the abstracted results. Because we were unable to calculate an effect size in nine of the 12 studies and because many of the studies with negative results probably lacked statistical power, we could not come to conclusive recommendations for these treatments.

Table 3

Trials of physical medicine methods for treatment of neck pain

View this table:

Treatments with positive results—In two strong quality randomised controlled trials the use of pulsed electromagnetic therapy (PEMT) produced a significant reduction in pain compared with placebo PEMT.20 21. Both studies used additional control treatments: one consisting of the use of a soft collar in combination with analgesic and anti-inflammatory drugs20 and the other the use of a soft collar in combination with education (advice), exercise, and analgesic and anti-inflammatory drugs.21 A combination of P values across these two studies produced an overall estimate of a positive effect of PEMT (P 0.0089; table 4. Two randomised controlled trials investigated acupuncture: one was a study of moderate quality that compared acupuncture with a placebo,38 and the other was a study of weak quality that compared acupuncture with a combination of traction and short wave diathermy.32 Both studies reported that the acupuncture group achieved better pain relief than the placebo group.

Table 4

Results of trials that used physical medicine methods with non-parametric data

View this table:

Treatments with mixed results—Two randomised controlled trials have been reported on the use of exercise; one with negative and the other with positive results.22 30 Goldie et al compared exercise in combination with drug treatment and advice to a control group in a study of moderate quality and reported no significant difference.22 Levoska et al compared active exercises with a combination of stretching, heat, and massage in a methodologically strong study and reported a significant difference (P</=0.01) favouring active exercise.30

Treatments with negative results—One methodologically strong randomised controlled trial compared spray and stretch (vapocoolant spray followed by passive stretching) with a placebo and found no significant difference between groups.40 Two trials, one methodologically strong42 and the other moderate,44 compared laser therapy with a placebo. Both report no significant difference between groups. P values from these studies were combined to give an overall estimate of effect of P = 0.6287 (table 4), indicating laser did not reduce pain. One randomised controlled trial of strong methodological quality compared infrared light therapy with a placebo.21 No significant reductions in pain were reported. Three randomised controlled trials investigated the use of traction.22 32 37 In a moderate quality study, Goldie et al compared traction with a control treatment of analgesics, muscle relaxants, and postural advice.22 They reported the difference between groups to be small and not significant, but no statistical analysis was reported. Pennie and Agambar, in a study of weak quality, compared traction, exercise, and patient education with collar and exercise.37 No significant difference between groups was reported, and no statistical analysis was described. A methodologically weak study by Loy compared traction and shortwave diathermy with electroacupuncture.32 Electroacupuncture was reported to improve symptoms significantly better than traction and diathermy, but no details of the analysis were reported. The one randomised controlled trial of moderate quality that compared TENS with a combination of collar, rest, education, and analgesics had an effect size of −0.5 (−1.9 to 0.4). We can be 95% certain that there was no difference between these treatments.36

Drug treatments

Four randomised controlled trials that used drug treatments met our eligibility criteria, with validity scores varying from strong to weak.14 15 19 25 Table 5 summarises the abstracted results.

Table 5

Trials of drug treatment for neck pain

View this table:

Two placebo controlled trials, one of moderate methodological quality15 and the other strong,14 investigated muscle relaxants with algometry or muscle spasm symptom scores as outcomes. Direct pain score data could not be abstracted from either study. Both reported significant improvement in patients with both neck or low back disorders. One methodologically weak randomised controlled trial compared a topical anti-inflammatory applied with TENS with TENS alone and reported significant pain reduction with the combined treatment.19 We could not calculate the magnitude of this effect. One strong randomised controlled trial tested a combination of anti-inflammatories, analgesics, and patient education compared with a placebo of detuned electrotherapy25 and showed no significant difference between treatments, but the small sample size used in this subgroup analysis probably does not have sufficient power to state conclusively that no difference exists.

Patient education

Three randomised controlled trials that used patient education (including ergonomic advice, neck school, postural advice, and strategies for the management of pain) met our inclusion criteria, with validity scores varying from weak to strong.24 25 34 Table 6 summarises the abstracted results.

Table 6

Trials of patient education for treatment of neck pain

View this table:

In Kamwendo et al traditional neck school and neck school combined with ergonomic advice were each compared with a no treatment control in a moderate quality study.24 Data provided by the authors allowed us to calculate an effect size, which showed no significant difference between groups. Two randomised controlled trials investigated forms of patient advice. Koes et al, in a methodologically strong study, compared a combination of anti-inflammatories, analgesics, and patient education with a placebo and found no significant difference between groups.25 McKinney et al compared two forms of education (advice) in a methodologically weak study.34 At four weeks of treatment, education combined with posture, relaxation training, and exercise gave significant pain relief compared with advice and rest. At eight weeks of treatment this effect disappeared.


Adverse effects have not been well documented. If we exclude data from the three trials in which patients with neck pain were not separated from those with low back pain,14 15 25 1254 patients were randomised in 21 randomised controlled trials. Six trials reported a total of 16 patients with increased symptoms or side effects resulting from treatment. No serious complications or deaths were reported.


Current practice often lags behind available evidence. In a survey about attitudes on treatment of musculoskeletal disease, active exercise, traction, TENS, and ultrasound were perceived to be the best methods for the treatment of neck pain.45 The results of this overview clearly do not support these approaches nor others commonly used in practice today.7 8 What becomes most clear from this overview is the lack of evidence for many of the standard approaches to neck pain used in health care today. Even for the treatments found in this overview to have some early evidence of support, such as manual treatments in combination with other treatments, conclusions must be made cautiously because of the small number of trials on which they are based.

The growing magnitude and socioeconomic impact of neck pain in society will demand that more research be conducted into the efficacy and effectiveness of its treatment. As more studies are accumulated, the influences of a number of factors on the results can be explored. Methodological quality, which has been shown in previous work to be related to outcome,46 was not shown to influence the pooled estimate of effect for manual treatments in combination with other treatments. Other clinically important influences such as type or chronicity of the disorder may also be expected to influence the results of the pooled analysis in future. Regular updates through the Musculoskeletal Review Group of the Cochrane Collaboration will facilitate this.47

In general, no treatments have been studied in enough detail to assess either efficacy or effectiveness adequately. When we consider the varied treatment approaches to neck pain, with their potential risks and costs, much further work is needed to determine optimal treatment approaches.

We are indebted to the authors of the primary studies included in this review and to Dr Eldon Tunks, Tom Flemming, Dr Deborah Cook, and Dr Bart Koes for their assistance at various stages of this work.


  • Funding This study was supported in part by a grant from the McGregor Clinic Fund, Hamilton Foundation.

  • Conflict of interest None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
View Abstract