Effectiveness of Acupuncture for Low Back Pain: A Systematic Review

Jing Yuan, PhD; Nithima Purepong, MSc; Daniel Paul Kerr, PhD; Jongbae Park, KMD, PhD; Ian Bradbury, PhD; Suzanne McDonough, PhD


Spine. 2008;33(23):E887-E900. 

In This Article


Study Selection

In total, 1606 studies were found, and 40 potentially eligible RCTs were identified, 15 of which were excluded in the first step (Figure 1).

The QUORUM statement flow diagram.

Adequacy of Acupuncture Treatment

Data on acupuncture treatment details were extracted and summarized. In general, acupuncture treatment details, i.e., chosen points, number of points needled, needle sensation, needle retention time, treatment frequency, and treatment sessions, were generally in line with textbooks,[35,36,37,38,39] surveys,[40,41,42] and reviews.[43,44,45,46,47,48] The exception is that 2 RCTs provided only 1 treatment session in total for chronic LBP, which was considered inadequate and excluded from this review.[49,50]

Finally, 23 RCTs were included, and the process of study selection was shown by a flow diagram as recommended in the Quality of Reporting of Meta-Analysis (QUOROM) statement[51] (Figure 1).

Varied styles of acupuncture have been used in the included RCTs, i.e., individualized (52%), standardized (22%), and semistandardized (26%) acupuncture. Semistandardized acupuncture has been defined as a set formula of points, supplemented by some additional points individually chosen for each patient.

Study Characteristics

Twenty-three RCTs representing 6359 LBP patients were included, and their study characteristics are provided in Table 1 . The sample size ranged from 17 to 3093, where 9 studies (39%) included between 50 and 100 subjects and 10 studies (43%) included more than 100 subjects. Nineteen (83%) studies were on chronic LBP (≥12 weeks), 1 study on subacute LBP (≥4weeks and <12weeks), and 3 studies on chronic and subacute LBP.

All 23 studies measured pain intensity, using visual analogue scales (VAS), numerical rating scales (NRS), SF-36 bodily pain dimension, Von Korff chronic pain grading scale, or LBP rating scale. Sixteen (70%) studies measured functional disability. Furthermore, 9 studies (39%) measured range of motion (ROM), 11 (48%) measured analgesic intake, 8 (35%) measured general health status, and some included the measures such as global assessment (2 RCTs) and adverse effects (5 RCTs).

Eight studies (35%) only had short-term follow-up, 12 (52%) intermediate term, and only 3 (13%) long-term follow-up. Thirteen studies (57%) had dropout rates less than 20% and 30% for short-/intermediate and long-term follow-up, respectively. Fourteen studies used follow-up interview 43% of them with large dropouts, and 9 studies used telephone/mail follow-up with 22% of them with large dropouts, which seemed superior over interview.

Thirteen studies did not account for missing data, whereas 10 studies (43%) adopted intention-to-treat analysis (ITT), of which 2 studies had no dropout,[17,52] 4 carried baseline,[53] discharge,[18] or last values forward,[11,54] 2 counted the missing data as failures/successes,[21,55] and 2 studies did not specify their ITT methodology.[56,57] However, no relationship could be explored between the analytic methods and the results.

Methodologic Quality Assessment

In summary, although 16/23 of the studies (70%) scored highly on the Van Tulder scale, only 8/23 had more than 40 patients per group of which 2 studies had high dropouts,[55,58] leaving only 6/23 high quality studies.[11,18,20,21,56,59]

Best Evidence Synthesis

In total 5 types of comparisons were made as below.

Acupuncture Versus No Treatment (n = 3). One high[18] and 2 low quality studies[60,61] provided moderate evidence that acupuncture was more effective than no treatment for short-term pain relief and conflicting evidence for intermediate pain relief.[60] There was moderate evidence for such a comparison for short-term functional improvement[18] ( Table 1 , Table 2 ).

Acupuncture Versus Sham Interventions (n = 8).

  1. Acupuncture versus sham acupuncture (n = 4): 3 high-quality studies provided strong evidence of no significant difference between acupuncture and sham acupuncture, for short-term and intermediate pain relief and functional improvement (n = 298 and n = 1162, respectively, using superficial needle insertion at nonacupoints without stimulation as sham acupuncture),[18,21] or for pain relief during and at the end of treatment (n = 190, cross-over design using superficial needle insertion with 2% lidocaine injection as sham acupuncture).[59] Although 1 low-quality study showed trigger point acupuncture was significantly superior over sham acupuncture (nonpenetrating) for pain and functional improvement at short-term follow-up, such a conclusion was unreliable given its small sample size (n = 26).[19]

  2. Acupuncture versus placebo transcutaneous electrical nerve stimulation (TENS) (n = 4): two low-quality studies showed no significant difference for pain relief between acupuncture and placebo TENS at discharge[57,62] and intermediate follow-up.[57] However, the conclusion is unreliable because both of them included less than 40 patients per group and had large dropouts. The study by Lehmann et al[57] also lacked between-group statistical comparisons. In contrast, the other 2 low-quality studies showed significant superior effects of acupuncture over placebo TENS for short/intermediate term pain relief.[52,63] However, their results were also unreliable because 1 study had high dropouts and both had less than 40 patients per group[63] ( Table 1 , Table 2 ).

Acupuncture Versus Conventional Therapy (n = 6). In this review, conventional therapy was defined as any other therapy except acupuncture, e.g., standard GP care including medication, physiotherapy (PT) etc. As a result, 6 studies provided conflicting evidence.

Acupuncture was significantly superior, over conventional therapy for pain and functional improvement at short/intermediate term follow-up in 1 high-quality study,[21] or over TENS for pain relief at discharge in 1 low-quality study, which was, however, unreliable due to the very small sample size (n = 20).[64] Two low-quality studies found no significant difference between acupuncture and TENS,[57,65] which was also unreliable due to the small sample size and lack of between-group statistical comparisons in both studies, and high dropouts.[57]

One high-quality RCT (n = 262) concluded that there was no difference between massage and acupuncture for pain relief at discharge,[56] but massage was more effective than acupuncture for pain relief at long-term follow-up. In terms of disability at short-term follow-up, massage was significantly more effective than acupuncture; however, at long-term follow-up, this difference was only marginally significant (P = 0.05). Moreover, there was no significant difference between acupuncture and self-care for pain and functional disability at short/long-term follow-up.

Two low-quality studies concluded that chiropractic spinal manipulation was more effective than acupuncture,[66,67] for pain and functional improvement, at discharge.[66,67] However, both studies included less than 40 patients per group, did not report between-group statistical comparisons, and 1 study had a high dropout,[67] all of which makes the evidence unreliable ( Table 1 , Table 2 ).

Acupuncture and Conventional Therapy Versus Conventional Therapy (n = 8). Two high-quality studies[11,20] and 5 low-quality studies[17,53,55,58,68] provided strong evidence that acupuncture combined with conventional therapy was more effective than conventional therapy alone for pain relief, and moderate evidence for functional disability,[11,53,54,58] at discharge or short-term/intermediate/long-term follow-up, respectively. Seven studies got high Van Tulder scores, but 3 of them had less than 40 patients per group,[17,53,54] and the other 2 had high dropouts at the intermediate follow-up,[55,58] despite both including group sizes of more than 40 patients ( Table 1 , Table2 ).

Acupuncture and Conventional Therapy Versus Sham Acupuncture and Conventional Therapy (n = 2). Two low-quality studies with high Van Tulder scores, more than 40 patients per group but large dropouts at intermediate follow-up, provided conflicting and unreliable evidence: 1 study (n = 126)[55] showed significant superior effects of acupuncture plus PT over sham acupuncture plus PT, on pain relief at discharge and intermediate follow-up. The other study (n = 100)[58] reported that acupuncture plus PT did not improve pain and function significantly compared with sham acupuncture plus PT at short/intermediate term follow-up ( Table 1 , Table 2 ).

10/31 studies for pain (31 comparisons) and 9/26 studies for functional disability (26 comparisons) provided sufficient data for calculation of effect sizes for these respective outcomes. With regards to both pain and functional disability, in general, moderate to large effect sizes have been achieved in the comparison of acupuncture versus no treatment,[18] or acupuncture plus conventional therapy versus conventional therapy alone,[11,17,53,55,58] whereas other groups of comparisons generally achieved small to moderate effect sizes (Figures 2, 3).

SMD of pain.

SMD of functional disability.

Clinical Significance

The mean differences for functional disability could only be calculated from a few studies, therefore it was considered insufficient to judge the clinical significance of this outcome.

Fortunately, all of the included 23 studies measured pain intensity, 12 of which provided sufficient data for the calculation of mean difference between groups, 7 studies used VAS (0-100 mm), 3 used NRS (0-100 mm), 1 used Short Form-36 (SF-36) bodily pain dimension (0%-100%), and 1 used Von Korff Chronic Pain Grading Scale (0-10). All of the 12 studies (33 comparisons) favored acupuncture in terms of pain reduction. Twenty-four percent (8/33) of comparisons achieved the MCID (-20% or more) on pain reduction;[17,18,19,55,61,64] however, only 2 of them clearly achieved the MCID, i.e., both limits of 95% CI of mean difference were greater than the MCID[19,55] (Figure 4).

Mean difference (95% CI) of pain on VAS/NRS/SF-36 bodily pain/Von Korff CPGS (100%). Zero: as indicated by the upper solid line, suggests no difference between treatment and control group. Positive estimates favor control group; negative estimates favor acupuncture group. MCID (minimal clinically important difference, -20%): as indicated by the lower dashed line, suggests that values of the between-group changes greater than 20% (below the dashed line) are clinically significant.
AT = acupuncture; CT = conventional therapy; N/A = not available; VAS = visual analogue scale; CPGS = chronic pain grade scale; NRS = numerical rating scale; SF-36 = short form 36; Follow-up = follow the patients from the end of treatment.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.