We defined long term as the time point after 9 months that was closest to 12 months ( van Tulder et al 2003). Data were

extracted by the lead author (AML) and by a second reviewer working independently (KMR, CGM, JHMc). For trials with continuous outcomes the mean, standard deviation, and sample size of follow-up scores or change from baseline scores were extracted. If not reported, means and standard deviations were imputed from the reported measures of central MLN0128 tendency and variance (Higgins and Green 2006). For trials with dichotomous outcomes the number of subjects experiencing the outcome of interest and the total sample size were extracted. Where continuous outcomes were reported in an individual study, the effects of the intervention were expressed as a mean difference with a 95% CI for each outcome. Where pooling of outcomes was deemed appropriate, a metaanalysis was conducted using a random effects model and the results were expressed as weighted mean differences. Pain and disability scores were converted to a 0–100

point scale prior Dactolisib concentration to calculation of effect size to enable comparison of outcomes between interventions and trials. Where dichotomous outcomes were reported, the effects of the intervention were expressed as the relative risk of beneficial outcome with 95% CI. From 24 419 titles identified by the searches, 254 full-text publications were retrieved, of which 33 were included in the review. (Reasons for exclusion are presented in Figure 1.) Quality: Trial quality was generally high with 60%

of trials scoring at least 7 out of 10 on the PEDro scale ( Table 1). The quality criteria related to blinding were commonly not met, with 17 trials not blinding participants and 26 trials not blinding therapists. Some of the interventions investigated, such as neck manipulation and exercise, are difficult to deliver with adequate blinding of participants or therapists. The other quality criteria that were most commonly not met were intention-to-treat analysis (22 trials) and concealment of treatment allocation (15 trials). Participants: The majority of the eligible trials investigated participants with chronic neck pain (n = 19) or neck pain of mixed duration (n = 11). A single eligible trial next ( Pikula 1999) investigated acute neck pain. Two trials did not specify the duration of the episode of neck pain. (See Table 2.) Interventions: The types of interventions investigated by the included trials were medications, relaxation, acupuncture, exercise, manual therapy, multi-modal intervention, and electrotherapy. (Specific interventions are presented in Table 2.) No eligible trials investigated the role of surgery, injections, or radiofrequency neurotomy for non-specific neck pain. The control intervention was a sham physical intervention in 20 trials, minimal intervention in 8 trials, no intervention in 3 trials, and placebo medication in 2 trials.