Chronic low back pain (CLBP) is a major health issue in the western world and is a significant burden on health care; Americans spend $37 billion annually with a further $19.8 billion lost in absenteeism . There is 58% life time prevalence of back pain in the UK, a 22-65% 1-year prevalence and 6-7% of all adults have constant back problems .
Although CLBP is usually benign (<1% serious pathology ) yet there are many therapeutic approaches. Acupuncture has become a popular treatment but its effectiveness has not been sufficiently demonstrated.
Modern (Verum) acupuncture originates in ancient Chinese philosophy which claims pain and disease manifest because of imbalances in bodies forces of Yin and Yang. It is believed these forces flow through specific courses (meridians) and can be manipulated using specific acupuncture points to regain the balance. Acupuncture has evolved from the traditional Chinese application and some styles incorporating adjuncts such as electrical stimulation of the acupuncture needle ,
A recent systematic review of articles published between 1966 and February 2003  concluded that the efficacy of acupuncture on CLBP was inconclusive due to the low methodological quality of selected studies. They found acupuncture had some short-term improvements in pain and function compared to control or sham but due to low methodological quality they concluded a need for higher quality studies.
This review updates that study  by including articles published after February 2003 or studies that were published prior but were of high relevance and methodological quality. The objective is to provide firm conclusions about the efficacy of acupuncture therapy for CLBP.
Only randomised controlled trials (RCTs) available in English and available free of charge were included.
In October 2009 the MEDLINE database (period 1950 to date) was searched for RCTs published after February 2003 and matching the search string “Chronic low back pain” AND “acupuncture” OR “dry needling” OR “Sham” OR “Placebo” AND “randomised controlled trial” OR “randomized controlled trial”. Further searches using PEDro, Web of Science (using ISI Web of Knowledge) and Cinahl (period 1982 to date) (see appendix A). Each articles reference list was also used as a source of relevant publications.
For inclusion the studies participants needed to be =18-years old with non-specific CLBP. Non-specific CLBP was defined as pain between the 12th costal margin and the inferior gluteal folds =12-weeks. If radiating leg pain was present this must be secondary to the lumbosacral region pain. RCTs that included participants with specific pathologies as the root cause of their CLBP, such as malignancy, prolapse of =1 inter-vertebral disc or spinal fracture were excluded.
Studies that investigated the effects of traditional (Verum) acupuncture, trigger-point acupuncture and dry needling were reviewed. RCTs were included regardless of hand of electro-stimulation. Studies investigating non-needle based acupuncture, such as laser acupuncture, were excluded. Control interventions included sham, usual care, Transcutaneous Electrical Nerve Stimulation (TENS) or conservative orthopaedic therapy.
There are four outcome measures considered to be important when assessing CLBP
RCTs must include =1 of the above.
The primary outcomes were pain and function.
A total of 544 studies were found through the searches with 17 potentially eligible RCTs identified. Of these 5 were excluded due to study duplication (n=1), sole inclusion of participants with specific CLBP (n=2) or use of non-needle based acupuncture (n=2).
The remaining 12 articles were reviewed using the Critical Appraisal Skills Programme (CASP) to determine their methodological quality. CASP enables the systematic review of an RCT for validity, design, execution and reasoning. Assessment criteria included randomization and allocation of participants, blinding of participants and assessors, identification of potential observer bias, participant numbers at RCT start and conclusion, presentation and accuracy of results, and any identified limitations. Results were recorded and documented (Appendix B).
 298 participants with CLBP =6-months randomised to 12 sessions of acupuncture (n=146) or sham acupuncture (n=73) over 8-weeks, administering therapists had =140 hours training and 3-years experience, with a third delayed acupuncture group (n=79) who received no acupuncture for the initial 8-weeks followed by the acupuncture groups protocol. Outcome measures were VAS-P and back function using the validated German Funktionsfragebogen Hannover-Rücken (FFbH-R) questionnaire. At 8-weeks VAS-P decreased from baseline in all groups; after 26 and 52-weeks the acupuncture group’s results were better than sham however differences were not significant. Results from the delayed acupuncture group followed the acupuncture groups’ pattern. The trial had good methodological quality: outcome measures were assessed independently with participants completing questionnaires, attrition was reasonable (18%) but the acupuncture group was double the size of the others which may have influenced results.
 638 participants with CLBP =3-months randomised to standard acupuncture (n=185), individualised acupuncture (n=157), sham acupuncture (n=162) or usual care (161) groups. Acupuncture groups received 10 treatments over 7-weeks by acupuncturists with =3 training. The Primary outcome measure was RMDQ. Compared to baseline all groups showed improved function and pain at 8-weeks. Mean values for RMDQ were consistent up to 52-weeks with the usual care group having greater dysfunction than all acupuncture groups (P=.001). There was no significant difference between real and sham acupuncture groups (P>0.05). All forms of acupuncture had beneficial and persisting effects over usual care for CLBP treatment with clinically meaningful functional improvements. There were no significant differences between acupuncture groups. Outcome measures were gathered by blinded telephone interviewers and attrition was low (6%) resulting in good trial internal validity
 1162 participants with CLBP =6-months randomised to 5-weeks of twice-weekly acupuncture (n=387) or sham acupuncture (n=387), performed by acupuncturists with =140 hours training. A third group received usual care (n=387). Outcome measures were Von Korff Chronic Pain Grade Scale (GCPS) and Hanover Functional Ability Questionnaire (HFAQ). Results were presented as a percentage of improvement in function and pain at 6-month follow-up. At 6-months both acupuncture groups had significant improvements in pain and function compared to baseline and usual treatment. There was no difference between acupuncture groups (p=0.39). The trial was methodologically strong with good internal validity: the control group was an active multimodal conventional therapy, had high power with stated calculation, follow-ups at 1.5, 3 and 6-months, low attrition (4%) and balanced dynamic randomisation. This was a good, highly relevant, large, rigorous trial.
 35 participants, =65-years, with CLBP =6-months randomised to 1 of 3 groups receiving 2 3-week phases of 30-minute acupuncture sessions, with a 3-week interval between. Group A (n=12) received standard acupuncture, Group B (n=10) superficial trigger-point acupuncture and Group C (n=13) deep trigger-point acupuncture. Outcome measures were VAS-P and RMDQ score. Group C showed a statistically significant VAS-P and RMDQ reductions from baseline after phase 1 with VAS-P reduction persisting over 12-weeks. There was no significant reduction in VAS-P or RMDQ for either other groups. The RCTs methods are described well however small sample size, high dropout (27%), short-term follow-up and potential bias limited internal validity
 26 participants, =65-years, with CLBP =6-months randomised to 2 groups. Over 12-weeks each group received 1 phase of trigger-point acupuncture and 1 phase of sham acupuncture with a 3-week break between. Group A (n=13) received trigger-point phase first followed by sham, Group B (n=13) vice-versa. Acupuncturist had =4-years training and =7-years clinical experience. Outcome measures were VAS-P and RMDQ score. After phase 1 Group A had significantly lower VAS-P (P<0.001) and RMDQ scores (P<0.001) than group B. Both groups showed significant within-group decreases in VAS-P (P<0.001) and RMDQ scores after trigger-point acupuncture but not during sham. At 12-weeks effects were not sustained. This trial was described well but suffers from the same shortcomings as the 2004 RCT by the same authors.
 60 participants with CLBP =6-months randomised them to 6-weeks of 30-minute weekly sessions of either acupuncture (n=30) or placebo TENS (n=30). No details of administering therapists were given. The primary outcome measure was VAS-P. Although acupuncture showed highly significant differences in all the outcome measures between pre- and post-treatment, the differences between the 2 groups were not statistically significant. Generally the RCT was poor: therapists were not blinded, high noncompliance (23.3%), cointerventions might have influenced results, the dropout rate was not explained and there was no intention-to-treat analysis.
 131 participants 18-65 years old with CLBP =6-months were randomised to groups receiving 20 30-minute sessions of traditional and auricular acupuncture (n=40), physiotherapy (n=46) or sham acupuncture and physiotherapy (n=45), over 12-weeks. Outcome measures were VAS-P and pain disability index (PDI). After 12-weeks of treatment the acupuncture group showed significantly reduced pain and disability compared to the physiotherapy group but not compared to the sham group. At 9-months the acupuncture group was more effective than physiotherapy in reducing disability only and not different to sham. The trial was methodologically strong but short-term dropout was 24% and long-term 37%. The treatment scheduled was five-a-week for 2-weeks then weekly for 10-weeks which may not be clinically practical.
 55 participants =60 yrs, with CLBP =12-weeks were randomised to 2-weeks of twice-weekly acupuncture and electrical stimulation alongside usual care (n=31) or usual care alone (n=24). Primary outcome was RMDQ. At 6-weeks results indicate clinically and statistically significant improvements in the acupuncture group for pain and disability compared to control. Effects remained and only diminished slightly at 9-weeks follow-up. The trial was methodologically strong: balanced randomisation, clear methods, low attrition (14%). Participant inclusion criteria included prior imaging limiting generalisability.
 186 participants aged between 20 and 60 with CLBP =6-week were randomised to 4-weeks of usual care alone (n=60) or with either acupuncture (n=65) or sham acupuncture (n=61). Acupuncturists were experienced doctors trained in Beijing. Primary outcome measure was VAS-P. Immediately after treatment 65% of the acupuncture group reported a =50% reduction in VAS-P compared to 34% of the sham group and 43% of the usual care group. At 3-months 79% of the acupuncture group, 29% of the sham group and 14% of the usual care group reported a =50% VAS-P reduction. Methodological quality was high: balanced (stratified) randomisation and excellently described methods however there was 30% attrition at 3-month follow-up and data collection was from general practitioners leading to potential performance bias.
 241 participants, aged 18-65, with CLBP for 4 to 52-weeks were randomised to 10 sessions of acupuncture (n=160) or to usual care (n=81) over 3-months. Acupuncturists were training for =3-years and =12.8-years clinical practice. Outcome measures were SF-36 pain scores and Oswestry low back pain disability questionnaire (ODI) taken at baseline, 3, 12 and 24-months. A power calculation stated a required 100 participants per group to detect a 10-point difference on SF-36 (90% power and 5% significance level). A 5 point difference in SF-36 was deemed significant. The number of participants in the acupuncture group was increased to 160 to allow for between-acupuncturist effect, usual care group decreased to 80 participants without power loss. Results were presented as point differences between randomisation, 12 and 24-months. At 12-months a 5.6 point intervention effect difference in SF-36 pain was found and 8 point at 24-months. No treatment effect was found for any other dimension of SF-36 or ODI. Participants were representative of UK population, randomisation was balanced, methods were thoroughly documented and acupuncture treatments were individualised resulting in high methodological quality and generalisability. However 25% of participants were unaccounted for at conclusion reducing internal validity.
 11630 participants with CLBP =6-months were allocated to three groups. Group A were received 15 individualised acupuncture sessions with usual care as needed (n=1549). Group B received delayed acupuncture with usual care as needed (n=1544). Group C declined to be randomised but received 15 individualised acupuncture sessions with usual care (n=8004). Treatment was over 3-months. Outcome measures were FFbH-R and SF-36 pain scores. At 6-months the acupuncture group showed significant improvements in FFbH-R and SF-36 pain compared to routine care alone. The large sample size and broad inclusion criteria meant results were generalisable however groups were different at baseline and findings identified a degree of randomisation selection.
 52 participants with CLBP =6-months were randomised to 4-weeks of physiotherapy with daily 1-hour electro-acupuncture sessions (n=26) or standard physiotherapy (n=26). Outcome measures were pain (NRS-P) and function using the Aberdeen-LBP. There was a significant reduction in NRS-P and Aberdeen-LBP scores in the acupuncture group immediately after treatment and at 1 and 3-months follow-up. Methodological quality was limited by possible breach of blinding integrity due to lack of patient blinding and subjective outcome measures.
Two high quality studies (11928 people)  and  found acupuncture more effective in short-term pain reduction and functional improvements than no (delayed) treatment. However both studies were weakened by insufficient blinding and participants were recruited from newspaper adverts  or an insurance company  limiting generalisability; both of which reduce results confidence.
Studies comparing acupuncture and sham acupuncture (2460 people) (, , ,  and ) found both effective at reducing pain and increasing function compared to baseline measures; however no study found a clinically significant difference between groups With five methodologically sound trials all reporting similar results clinicians can have confidence in the effectiveness of acupuncture or sham-acupuncture in pain and functional improvements. However with no clinically significant difference between groups, placebo effect seems to be a substantial contributing factor.
Five RCTs comparing acupuncture and usual care (12164 people) (, , ,  and ) concluded that acupuncture was more effective at reducing pain. Increased function in the acupuncture group compared to control was reported in 1 RCT  at 6- and 9-weeks,  at 6-month and another  investigated effect immediately after treatment and 1- and 3-months follow-up; however 1 RCT  found no significant improvement in function in their longer-term study at 12 or 24-month. Unlike other papers reviewed, Thomas and colleagues used UK based participants who received treatments in private or GP clinics adding confidence to their conclusions when applied to the general UK population. From study findings clinicians can have confidence that the addition of acupuncture to their treatment of CLBP will be more effective than usual care alone.
One study ( 35 people) found greatest improvements in pain and function using deep trigger-point-acupuncture. However this study, while being methodologically thorough and having patient and assessor blinding, was limited by small size, high dropout (23%), short-term follow-up and possible centre bias leading to reduced clinical confidence.
One RCT ( 60 people) found significant improvements using both TENS and acupuncture but no significant intergroup difference over 6-months. However, confidence in results are limited because participants also received usual care and exercise so may have improved regardless; furthermore the study had no therapists blinding, high noncompliance (23.3%), unexplained dropouts and no intention-to-treat analysis.
In 1 cross-over trial ( 26 people) trigger-point acupuncture was found to be more effective than sham however small sample size, high attrition (23%), restricted to short-term follow-up and possible bias due to centre location (Department of Orthopaedic Surgery, Meiji University of Oriental Medicine) limit confidence in findings.
Studies were commonly limited by being unrepresentative: of the 12 studies 2 were UK based (, ), six restricted participants by age (, ,  , , ), 2 used participant recruitment methods which may have introduced expectation bias (newspaper adverts, , insurance company ) and five had underpowered sample sizes or non-stated power calculations (, , , , ). Without representative sample groups the outcome measures cannot be applied to the general population with any reliability.
Discrepancies were noted in treatment frequency with control group participants receiving less attention than intervention participants .
Blinding was inconsistent across studies: 1 study () blinded participants in the acupuncture groups but not the delayed group, 1 study () blinded participants only, four (, , , ) blinded assessors and participant, 1 () blinded assessors only, 1 () blinded assessors and participants but not acupuncturists, three (, , ) had no blinding and 1 () blinded assessors but not participants.
There is some evidence for the efficacy of acupuncture for CLBP; compared to no treatment there was short-term ( 8-week and  3-month) pain reduction and functional improvements. Compared to sham therapy both showed similar improvements in pain and function at short-term ( 8-week,  8-week,  12-week and  3-month) and mid-term ( 6-month and 1-year,  6-month,  9-month) follow-up but no significant difference was detected between groups. Compared to usual care acupuncture showed significant improvements in primary outcome measures at treatment, short- (6- and 9-week, 3-month, 1- and 3-month) and long-term (6-month, 1- and 2-year) follow-up. Compared to superficial and deep trigger-point all treatments showed improvements but none were significantly different from each other. Both acupuncture and TENS were found to produce long-term ( 6-month) improvements but no significant difference was found between interventions. Comparing trigger-point therapy to sham, trigger-point was found to be more effective although benefits were not sustained. There is evidence that acupuncture alongside other treatments relieves pain and increases function better than individual therapies alone.
Further research needs to be conducted to determine treatment frequencies and sustainability of treatment effects. Effective sham treatments need to be developed to establish placebo effect compared to acupuncture and other therapy types.