Key Points

• For acute low back pain, one good-quality trial found no difference between acetaminophen versus placebo in pain intensity or function through 3 weeks (strength of evidence [SOE]: low for pain and function).
• For acute low back pain, a systematic review found no difference between acetaminophen versus nonsteroidal anti-inflammatory drugs (NSAIDs) in pain intensity (3 trials, pooled standard mean difference (SMD) 0.21, 95% confidence interval [CI] −0.02 to 0.43) or likelihood of experiencing global improvement (3 trials, relative risk [RR] 0.81, 95% CI 0.58 to 1.14) at ≤3 weeks, though estimates favored NSAIDs and the estimate was imprecise (SOE: insufficient)
• For chronic low back pain, no study evaluated acetaminophen versus placebo, and there was insufficient evidence from one trial to determine effects of acetaminophen versus NSAIDs (SOE: insufficient).
• There was insufficient evidence from four trials to determine effects of acetaminophen versus other interventions (SOE: insufficient).
• No study evaluated acetaminophen for radicular low back pain.
• One trial found no difference between scheduled acetaminophen, as-needed acetaminophen, or placebo in risk of serious adverse events (∼1% in each group) and a systematic review found acetaminophen associated with lower risk of side effects versus NSAIDs (3 trials, RR 0.57, 95% CI 0.36 to 0.89) (SOE: moderate).

Detailed Synthesis

The APS/ACP review²⁹ included eight trials of acetaminophen (Appendix Tables E1, F1). One trial evaluated acetaminophen versus no treatment,³⁰ five trials included in a systematic review^{31, 32} evaluated acetaminophen versus various NSAIDs,^{30, 33-36} and three trials evaluated acetaminophen versus other interventions (amitriptyline,³⁷ electroacupuncture,³⁸ and manipulation, corset, or physical therapy³⁹). The sample size was 456 in one trial³⁹ and ranged from 40 to 70 in the others. One trial evaluated acetaminophen for chronic low back pain,³⁴ one mixed acute to chronic low back pain,³⁹ and the remainder acute low back pain. No trial specifically focused on patients with radiculopathy. Acetaminophen doses were 4 g/day in 3 trials,^33-35 3 g/day in 1 trial,³⁰ 2 g/day in one trial,³⁷ and unclear in 3 trials.^37-39 Duration of treatment ranged from 1 to 5 weeks. Four trials evaluated patients at 3 to 9 weeks after the completion of therapy^{30, 35, 38, 39} and the remainder evaluated patients at the end of therapy. Two trials^{34, 37} were classified as higher quality (based on meeting fewer than 6 of 11 validity criteria) and the remainder classified as lower quality. Methodological shortcomings included inadequate or unclear randomization and allocation concealment methods, unblinded design, and failure to avoid cointerventions. The APS/ACP review concluded that there was good evidence that acetaminophen was associated with moderate effects for acute and chronic low back pain, based primarily on evidence that acetaminophen and NSAIDs were associated with similar effectiveness in most trials, and trials that evaluated effects of acetaminophen for other pain conditions.

An update^{40, 41} to the systematic review³² of acetaminophen versus NSAIDs included one additional high-quality trial (n=371) of patients with acute pain (Table 1).⁴² Acetaminophen was compared against ibuprofen, a heat wrap, an unheated wrap, and placebo after a 4-day course of therapy. However, results were only reported for the comparisons of acetaminophen versus ibuprofen or heat wrap.

Table 1

Summary of systematic reviews of pharmacological treatments for low back pain.

We identified one additional good-quality trial (n=1643) of scheduled (∼4 g/day) or as-needed (up to 4 g/day) acetaminophen for up to 4 weeks versus placebo for acute low back pain (Table 2, Appendix Tables E2, F2).⁴³ Followup was conducted through 12 weeks.

Table 2

Characteristics and conclusions of included acetaminophen trials.

Key Points

• For acute low back pain, a systematic review found NSAIDs associated with greater improvement in pain intensity versus placebo (4 studies, weighted mean difference [WMD] −8.39, 95% CI −12.68 to −4.10; chi-square 3.47, p>0.1), but four trials found no clear effects on the likelihood of achieving significant pain relief. One subsequent trial was consistent with these findings. One trial found NSAIDs associated with better function versus placebo (SOE: moderate for pain, low for function).
• For chronic low back pain, a systematic review found NSAIDs associated with greater improvement in pain versus placebo (4 trials, WMD −12.40, 95% CI −15.53 to −9.26; chi-square 1.82, p>0.5); two trials found NSAIDs associated with greater improvement in function (SOE: moderate for pain, low for function).
• For radicular low back pain, a systematic review found no difference in pain intensity between NSAIDs versus placebo (2 trials, WMD −0.16, 95% CI −11.92 to 11.59, chi-square 7.25, p<0.01) (SOE: low).
• There was insufficient evidence from two trials of an NSAID plus another intervention (skeletal muscle relaxants or massage) versus the other intervention alone to determine effectiveness (SOE: insufficient).
• There was insufficient evidence from two trials to determine the effects of NSAIDs versus doloteffin or exercise (SOE: insufficient)
• A systematic review found that most trials of one NSAID versus another found no differences in pain relief in patients with acute low back pain (15 of 21 trials) or chronic low back pain (6 of 6 trials) (SOE: moderate).
• A systematic review found NSAIDs associated with more side effects versus placebo (10 trials, RR 1.35, 95% CI 1.09 to 1.68); COX-2-selective NSAIDs were associated with lower risk of side effects versus nonselective NSAIDs (4 trials; RR 0.83, 95% CI 0.70 to 0.99). Serious harms were rare. (SOE: moderate)

Detailed Synthesis

The APS/ACP review²⁹ included a good-quality systematic review with 51 trials of NSAIDs.^{31, 44} The review found nonselective NSAIDs for acute (6 trials) and chronic (1 trial) low back pain moderately more effective than placebo for outcomes related to pain and function. The APS/ACP review also found no evidence that any nonselective NSAID is superior to another for pain relief based on 24 trials, or when compared with other active interventions (e.g., other medications or passive physical modalities.) None of the trials in the systematic review evaluated a COX-2 selective NSAID.

We identified an updated version of the systematic review described above (Table 1, Appendix Tables E3, F3).⁴¹ It included 65 trials (total n=11,237) of NSAIDs versus placebo (16 trials), other active interventions (13 trials), or one type of NSAID versus another (33 trials), including five trials of COX-2 inhibitors (meloxicam, nimesulide, valdecoxib or etoricoxib) versus nonselective NSAIDs.^45-49 Of the COX-2-selective NSAIDs evaluated in the trials, the only one available in the United States is meloxicam. Eleven trials investigated diclofenac sodium, eight trials ibuprofen, seven trials piroxicam, seven trials diflunisal, four trials naproxen, and 23 trials evaluated other NSAIDs. Nine trials of NSAIDs versus acetaminophen are discussed in the acetaminophen section of this report. Four trials in the systematic review of NSAIDs plus vitamin B versus NSAIDs alone are outside the scope of this review and not discussed further. Of the studies in the systematic review, 37 were conducted in patients with acute low back pain, nine in patients with chronic low back pain, and the remainder in patients with mixed or unclear duration of pain. Six studies included only patients with sciatica, 25 included low back pain without sciatica and 34 evaluated a mixed population or did not specify whether or not patients had sciatica. Treatment schedules and doses varied across studies. Medications were taken 1 to 6 times per day, and doses varied widely (i.e., ibuprofen doses ranged from 800 to 2400 mg per day, diclofenac doses ranged from 48 to 150 mg per day). Duration of treatment ranged from 1 day to 12 weeks, and followup ranged from 2 days to 6 months. 28 studies were rated high-quality by the systematic review, based on meeting at least 6 of 11 Cochrane Back Review Group criteria; the other 37 were rated low quality. Common methodological shortcomings of the low-quality studies included inadequate details regarding randomization and allocation concealment methods and cointerventions.⁴¹

We identified three additional trials (n=54 to 193) not included in the systematic review of NSAIDs for acute (including recurrent)⁵⁰ or subacute^{51, 52} low back pain (Table 3; Appendix Tables E4, F4). One trial compared lornoxicam, diclofenac, and placebo,⁵⁰ one trial an NSAID plus deep tissue massage versus deep tissue massage alone,⁵¹ and one trial an NSAID (loxoprofen sodium, diclofenac sodium, or zaltoprofen) versus exercise.⁵² One trial was rated good quality,⁵² and two were rated fair quality.^{50, 51} Methodological shortcomings of the fair-quality trials included inadequate description of randomization, blinding and avoidance of cointerventions.

Table 3

Characteristics and conclusions of included NSAID trials.

Key Points

• For chronic low back pain, a systematic review found opioids associated with greater short-term improvement in pain scores (6 trials, SMD −0.43, 95% CI −0.52 to −0.33, I²=0.0%, for a mean difference of ∼1 point on a 0−10 pain scale) and function (four trials, SMD −0.26, 95% CI −0.37 to −0.15; I²=0.0%, for a mean difference of ∼1 point on the RDQ) versus placebo; three additional trials reported results consistent with the systematic review (SOE: moderate for pain and function).
• For chronic low back pain, a systematic review found tramadol associated with greater short-term pain relief versus placebo (5 trials, SMD −0.55, 95% CI −0.66 to −0.44, I²=86%, for a mean difference of 1 point or less on a 0-10 pain scale) and function (5 trials, SMD −0.18, 95% CI −0.29 to −0.07, I²=0%, for a mean difference of ∼1 point on the RDQ); two trials not included in the systematic review reported results consistent with the systematic review findings (SOE: moderate for pain and function).
• For subacute or chronic low back pain, a systematic review included two trials that found buprenorphine patches associated with greater short-term improvement in pain versus placebo patches; effects on function showed no clear effect or were unclearly reported (SOE: low for pain, insufficient for function).
• For chronic low back pain, three trials reported inconsistent effects of opioids versus NSAIDS for pain relief, one trial found no difference in function. (SOE: insufficient for pain and function).
• For acute low back pain, one trial found no significant differences between opioids versus acetaminophen in days to return to work; pain was not reported (SOE: insufficient).
• Four trials found no clear differences among different long-acting opioids in pain or function (SOE: moderate for pain and function).
• Six trials found no clear differences between long-acting versus short-acting opioids in pain relief. Although some trials found long-acting opioids associated with greater pain relief, patients randomized to long-acting opioids also received higher doses of opioids (SOE: low).
• Short-term use of opioids was associated with higher risk versus placebo of nausea, dizziness, constipation, vomiting, somnolence, and dry mouth; risks of opioids were higher in trials that did not use an enriched enrollment and withdrawal design (SOE: moderate). Trials were not designed to assess risks of overdose, abuse, and addiction, or long-term harms.

Detailed Synthesis

The APS/ACP review included nine trials of opioid analgesics for low back pain.²⁹ Sample sizes ranged from 36 to 683 patients. Three trials compared opioids versus placebo or acetaminophen, five trials compared sustained-release versus immediate-release opioid formulations, and two trials compared two different long-acting opioids. Only one trial assessed opioids for acute low back pain; the remainder evaluated opioids for subacute or chronic low back pain. Two trials were rated higher quality. Based on this evidence, the APS/ACP review found fair evidence that opioids are moderately more effective than placebo or no opioid for subacute or chronic low back pain, but insufficient evidence to determine effects for acute low back pain.

A recent, good-quality systematic review⁶⁹ of opioids for low back pain included 16 randomized controlled trials (RCTs) (reported in 15 publications) (Table 1; Appendix Tables E5, F5).^70-84 Sample sizes ranged from 55 to 981 patients. The opioids evaluated were tapentadol (1 trial), oxycodone (2 trials), long-acting oxycodone (1 trial), long-acting morphine (2 trials), extended-release hydromorphone (1 trial), extended-release oxymorphone (3 trials), combinations of oxycodone with naloxone or naltrexone (2 trials), tramadol or the combination of tramadol and acetaminophen (7 trials), and buprenorphine patches (2 trials). In many trials the dose of opioids was titrated to achieve pain relief; maximal doses ranged from 20 to 256 mg of morphine-equivalent doses per day. Tapentadol, morphine, oxymorphone, and hydromorphone were classified as “strong” opioids and analyzed together; tramadol and buprenorphine (a partial opioid agonist) were analyzed separately. Fourteen trials compared an opioid versus placebo and two trials compared an opioid versus an NSAID. The duration of treatment ranged from 2 weeks to 13 weeks following titration, and outcomes were assessed through the end of therapy in all trials. The systematic review assessed 13 trials as being at low risk of bias based on meeting ≥6 of 12 Cochrane Back Review Group criteria. Methodological shortcomings included high attrition, uncertain adherence to treatment, and uncertainty about blinding of outcome assessments. Five trials used the enriched enrollment and randomized withdrawal design described below.

We also included three trials (one higher quality⁸⁵ and two lower quality^{36, 86}) from the APS/ACP review that were not included in the systematic review and 4 additional newer trials (Table 4; Appendix Tables E8, F6).^87-90 Sample sizes ranged from 36 to 302 subjects. One trial evaluated patients with acute low back pain³⁶ and the others evaluated patients with subacute or chronic low back pain. The opioids evaluated were long-acting oxymorphone, combined oxycodone and naloxone, long-acting morphine plus oxycodone, short-acting oxycodone alone, oxycodone plus aspirin, and codeine. Two trials compared opioids versus placebo, one compared an opioid plus naproxen versus naproxen alone, and one compared opioids versus acetaminophen. Two newer trials compared tramadol plus acetaminophen versus placebo, and one compared long-acting hydrocodone versus placebo. The duration of treatment ranged from 15 days to 16 weeks, with outcomes assessed at the end of treatment. Two of the newer trials were rated good quality,^{87, 88} one fair quality,⁹⁰ and one poor quality.⁸⁹ Methodological shortcomings in the poor- and fair-quality trials included failure to describe adequate randomization methods, failure to report baseline differences, unblinded design, and high attrition. We also identified a post-hoc analysis of one of the trials included in the systematic review that stratified results according to presence of neuropathic pain.⁹¹

Table 4

Characteristics and conclusions of included opioid trials.

Of the 23 total trials, 8 employed an enriched enrollment and withdrawal design. In this design, all potential subjects receive the study drug for a period of time in a prerandomization, open label phase. Only those who benefit from the drug and tolerate side effects are then randomized to continue the active drug, or have it withdrawn and replaced with a placebo. Thus, every patient entering the trial has already demonstrated benefit from the opioid and been shown to be free of intolerable side effects at the time of randomization. This strategy can help reduce dropout rates following randomization and reduce the number of unresponsive subjects. However, it may also overestimate efficacy, and has been shown to underestimate adverse events.⁹²

Despite frequent use of the enriched enrollment and withdrawal design, dropout rates from the trials were high. Only 2 trials had a dropout rate of less than 20 percent,^{86, 90} and most had rates of 30 to 60 percent.^{70, 72-80, 82, 84, 87, 89} The most common reasons for dropout were discontinuation due to adverse events (more common in the opioid arms than placebo) or lack of effect (more common in the placebo arms).

Another limitation of all the trials was short duration. The longest trial was 16 weeks (18).⁸⁶ Many trials excluded patients with a history of substance abuse or depression, though these were groups that were more likely than others to receive opioids in clinical practice.⁹³ Seventeen trials were industry sponsored and all involved tramadol, new long-acting preparations of older drugs, or new drug combinations.

We also included findings from a good-quality comparative effectiveness review⁹⁴ of opioids for chronic noncancer pain that included three head-to-head trials of different long-acting opioids for low back pain^{85, 95, 96} and five trials of long-acting versus short-acting opioids,^{86, 97-100} and two other trials that evaluated comparisons among opioids.^{101, 102}

Key Points

• For acute low back pain, a systematic review found skeletal muscle relaxants superior to placebo for short-term pain relief (≥two-point or 30% improvement on a 0-10 VAS pain scale) after 2 to 4 days (4 trials; RR 1.25, 95% CI 1.12 to 1.41; I²=0%) and 5 to 7 days (3 trials; RR 1.72, 95% CI 1.32 to 2.22; I²=0%); a more recent, large (n=562) trial was consistent with the systematic review (SOE: moderate).
• For acute low back pain, a systematic review found no difference between a skeletal muscle relaxant plus an NSAID versus the NSAID alone in the likelihood of experiencing pain relief, though the estimate favored combination therapy (2 trials; RR 1.56, 95% CI 0.92 to 2.70; I²=84%); one other trial (n=197) also reported results that favored combination therapy (SOE: low).
• For chronic low back pain, evidence from three placebo-controlled trials was insufficient to determine effects, due to imprecision and inconsistent results (SOE: insufficient).
• Three trials in a systematic review found no differences in any outcome among different skeletal muscle relaxants for acute or chronic low back pain (SOE: low).
• A systematic review found skeletal muscle relaxants for acute low back pain associated with increased risk of any adverse event versus placebo (8 trials; RR 1.50, 95% CI 1.14 to 1.98) and increased risk of central nervous system events (primarily sedation) (8 trials; RR 2.04, 95% CI 1.23 to 3.37; I²=50%); one additional placebo-controlled trial was consistent with these findings (SOE: moderate).

Detailed Synthesis

The APS/ACP review²⁹ included a good-quality systematic review of skeletal muscle relaxants¹⁰³ with 22 studies.^104-125 Twelve trials compared a skeletal muscle relaxant versus placebo,^{104-107, 109, 112, 114-116, 118, 119, 121} four compared a skeletal muscle relaxant plus an NSAID versus an NSAID alone,^{110, 113, 123, 125} two compared a skeletal muscle relaxant versus another active treatment,^{117, 124} and three compared one skeletal muscle relaxant versus another.^{111, 120, 122} The skeletal muscle relaxants evaluated were tizanidine (4 to 12 mg/day; 7 trials), cyclobenzaprine (30 to 40 mg/day; 4 trials), oral (dose range 100 to 200 mg/day; 3 trials) or intravenous (single 60 mg dose;1 trial) orphenadrine, carisoprodol (1400 mg/day; 2 trials), chlorzoxazone (1500 mg/day; 1 trial), dantrolene (dose not reported; 1 trial), baclofen (30 to 40 mg/day;1 trial), pridinol (8 mg IM, then oral 4 mg/day; 1 trial), tolperisone (300 mg/day; 1 trial) and meprobamate 450 mg/day; 1 trial). Duration of treatment ranged from 4 to 21 days in 21 trials, except for one trial of single dose intravenous orphenadrine.¹¹⁸ Only three trials followed patients after treatment had been completed.^{106, 123, 125} Sample sizes ranged from 20 to 405 (median n=80). Eighteen of the trials enrolled patients with acute back pain and four^{106, 112, 120, 121} enrolled patients with chronic back pain. The review classified 17 trials^{104, 105, 107-109, 111-114, 116-123, 125} as high quality based on meeting at least 6 of 11 Cochrane Back Review criteria; the other five trials were classified as low quality.^{106, 110, 115, 120, 124} Methodological shortcomings in most trials included inadequate reporting of randomization and allocation concealment methods; low-quality trials also did not report attrition, had unbalanced groups at baseline, and/or did not conduct intention-to-treat analyses. The APS/ACP review concluded that there was good evidence for moderate effects of skeletal muscle relaxants versus placebo for acute low back pain, but insufficient (poor) evidence to determine effects for chronic low back pain.

We identified two fair-quality trials of skeletal muscle relaxants for acute or subacute back pain published since the APS/ACP review (Table 5; Appendix Tables E7, F7).^{126, 127} One trial evaluated carisoprodol 1000 mg/day (250 mg three times daily) versus placebo (n=562)¹²⁷ and the other (n=197) tizanidine 4 mg/day (2 mg twice daily) plus aceclofenac (an NSAID not available in the United States) 200 mg/day (100 mg twice daily) versus aceclofenac alone.¹²⁶ In both studies, duration of treatment was 7 days, with no post-treatment followup. Neither trial provided information regarding methods of randomization or allocation concealment, or methods of blinding of study personnel.

Table 5

Characteristics and conclusions of included skeletal muscle relaxant trials.

Key Points

• For acute low back pain, there was insufficient evidence from two trials with inconsistent results to determine effectiveness of benzodiazepines versus placebo (SOE: insufficient).
• For chronic low back pain, a systematic review included two trials that found tetrazepam associated with lower likelihood of no improvement in pain at 5-7 days (RR 0.82, 95% CI 0.72 to 0.94) and at 10 to 14 days (RR 0.71, 95% CI 0.54 to 0.93) versus placebo, and lower likelihood of no overall improvement at 10 to 14 days (RR 0.63, 95% CI 0.42 to 0.97) (SOE: low).
• For acute or subacute radicular pain, one trial found no difference between diazepam 5 mg twice daily for 5 days versus placebo in function at 1 week through 1 year, or other outcomes including analgesic use, return to work, or likelihood of surgery through 1 year of followup. Diazepam was associated with lower likelihood of experiencing ≥50% improvement in pain at 1 week (41% vs. 79%, RR 0.5, 95% CI 0.3 to 0.8) (SOE: low).
• For acute low back pain, there was insufficient evidence from two trials with inconsistent results to determine effects of benzodiazepines versus skeletal muscle relaxants (SOE: insufficient).
• For chronic low back pain, one trial found no difference between diazepam versus cyclobenzaprine in outcomes related to muscle spasm (SOE: low).
• A systematic review found central nervous system adverse events such as somnolence, fatigue, and lightheadedness were reported more frequently with benzodiazepines versus placebo, though harms were not reported well; no trial was designed to evaluated risks with long-term use of benzodiazepines such as addiction, abuse, or overdose (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included a systematic review of skeletal muscle relaxants for low back pain¹⁰³ that included eight trials of benzodiazepines.^{106, 128-134} The sample size was 152 in one trial;¹³³ and ranged from 30 to 80 in the other trials. Four trials compared a benzodiazepine versus placebo,^{106, 130-132} one trial compared a benzodiazepine plus physical therapy versus placebo plus physical therapy,¹³³ and three trials compared a benzodiazepine versus a skeletal muscle relaxant (carisoprodol,¹²⁸ cyclobenzaprine,¹⁰⁶ or tizanidine¹²⁹). One other trial evaluated a benzodiazepine versus drugs that are not available in the United States.¹³⁴ Four trials in the systematic review evaluated benzodiazepines for acute low back pain^{128, 129, 131, 132} and three for chronic low back pain.^{106, 130, 133} Two trials specifically enrolled patients with muscle spasms.^{106, 129} No trial focused on patients with radiculopathy; in one trial the proportion of patients with radiculopathy was 40 percent.¹³¹ Five trials evaluated diazepam^{106, 128, 129, 131, 132} and two trials evaluated tetrazepam (not available in the United States).^{130, 133} Diazepam was administered orally at doses of 5 to 10 mg three or four times daily in three trials;^{106, 128, 129} two trials evaluated regimens that included initial intramuscular diazepam and oral doses.^{131, 132} In both trials of tetrazepam, the dose was 50 mg by mouth three times daily.^{130, 133} The duration of therapy ranged from 6 to 14 days; two trials^{106, 130} evaluated patients 4 days after the completion of therapy and the others evaluated patients at the end of therapy. The review classified five trials as high quality^{128-131, 133} based on meeting at least 6 of 11 Cochrane Back Review group criteria; the other two trials were classified as low quality.^{106, 132} All trials used a blinded design. Methodological shortcomings included inadequate reporting of randomization and allocation concealment methods; some trials also did not report attrition or intention-to-treat analyses. The APS/ACP review concluded that there was fair evidence of a moderate effect of benzodiazepines for acute and chronic low back pain, based in part on evidence from populations with mixed back and neck pain, but noted that there were no reliable data on the risk of abuse of addiction.

We identified one good-quality trial (n=60) published since the APS/ACP review of diazepam 5 mg three times daily for 5 days versus placebo for acute radiculopathy due to prolapse lumbar disc (with computed tomography scan or magnetic resonance imaging confirmation) (Table 6; Appendix Tables E8, F8).¹³⁵ Outcomes were evaluated through 1 year.

Table 6

Characteristics and conclusions of included benzodiapine trials.

Key Points

• For chronic low back pain, a systematic review found no differences in pain between tricyclic antidepressants versus placebo (4 trials; SMD −0.10, 95% CI −0.51 to 0.31; I²=32%) or SSRIs versus placebo (3 trials; SMD 0.11, 95% CI −0.17 to 0.39; I²=0%); there was also no difference between antidepressants versus placebo in function (2 trials, SMD −0.06, 95% CI −0.40 to 0.29; I²=0%) (SOE: moderate for pain, low for function).
• For chronic pain, three trials, found duloxetine associated with lower pain intensity (differences 0.58 to 0.74 on a 0 to 10 scale) and better function (differences 0.58 to 0.74 on the Brief Pain Inventory-Interference scale) versus placebo (SOE: moderate for pain and function).
• No study evaluated the effectiveness of antidepressants specifically for radicular low back pain.
• No study compared duloxetine versus a tricyclic antidepressant.
• Antidepressants were associated with higher risk of any adverse events compared with placebo, with no difference in risk of serious adverse events (SOE: moderate).

Detailed Synthesis

The APS/ACP review²⁹ included three higher-quality systematic reviews^136-138 of antidepressants for low back pain. The reviews included a total of 10 unique trials (8 placebo controlled). Based on the systematic reviews, the APS/ACP review²⁹ concluded that that tricyclic antidepressants were slightly more effective than placebo for pain relief for chronic back pain, with no significant effects on function. There was insufficient evidence to determine the effectiveness of antidepressants for acute low back pain.

We identified a good-quality systematic review¹³⁹ on antidepressants for low back pain published subsequent to the APS/ACP review (Table 1; Appendix Tables E9, F9). The review included 10 trials (n=16 to 121);^140-149 seven of the trials were included in one or more of the older systematic reviews.^{140, 142-145, 147, 148} Only two trials required patients to have depression in addition to low back pain;^{144, 148} the other trials excluded patients with depression,^{141-143, 146} patients with depression accounted for a minority of enrollees,^{140, 145, 149} or depression status was not reported.¹⁴⁷ The antidepressants assessed were tricyclic antidepressants (desipramine [3 trials], imipramine [2 trials], amitriptyline and nortriptyline [1 trial each]), selective serotonin reuptake inhibitors (paroxetine [3 trials], fluoxetine [2 trials]), tetracyclic antidepressants (maprotiline, trazodone [1 trial each]), and bupropion (1 trial.) One trial¹⁴⁹ evaluated injectable clomipramine, an intervention which is not widely used and outside the scope of this review. All studies included a placebo arm, though three used an active placebo (either diphenhydramine,¹⁴² benztropine,¹⁴¹ or atropine¹⁴⁸) meant to mimic the side effects of antidepressants without therapeutic effects on pain. One trial also compared a tetracyclic antidepressant (maprotiline) versus an SSRI (paroxetine).¹⁴² Duration of followup in the trials ranged from 10 days to 12 weeks. Nine of the trials enrolled participants with chronic pain; duration of pain was not reported in the other trial.¹⁴⁰ Seven trials were assessed as high quality based on meeting at least 6 Cochrane Back Review Group criteria. Methodological limitations in the three lower-quality trials included inadequate description of randomization and allocation concealment methods and high rates of attrition.¹³⁹

We identified five additional trials (n=60 to 404) not included in the prior systematic reviews of antidepressants for chronic low back pain (Table 7; Appendix Tables E10, F10).^150-154 Three trials compared duloxetine (a serotonin-norepinephrine reuptake inhibitor) versus placebo,^152-154 one trial duloxetine versus escitalopram,¹⁵¹ and one trial amitriptyline versus bupropion.¹⁵⁰ One trial was rated good quality,¹⁵² three trials fair quality,^{151, 153, 154} and one trial poor quality.¹⁵⁰ Methodological shortcomings in the poor- and fair-quality trials included inadequate description of randomization, allocation concealment, and blinding methods.

Table 7

Characteristics and conclusions of included antidepressant trials.

Key Points

• No trial evaluated antiseizure medications for acute nonradicular low back pain.
• One trial found no difference between gabapentin (up to 3600 mg/day) versus placebo for chronic nonradicular low back pain, but did not meet inclusion criteria because it was only published as an abstract (SOE: insufficient).
• For chronic radicular low back pain, there was insufficient evidence from three poor-quality trials with inconsistent findings to determine effects of gabapentin versus placebo (SOE: insufficient).
• For chronic radicular low back pain or mixed radicular and nonradicular low back pain, two trials reported inconsistent results for effects of topiramate versus placebo (SOE: insufficient).
• For chronic radicular low back pain, two trials reported inconsistent effects of pregabalin versus placebo for pain or function (SOE: insufficient).
• For chronic radicular or nonradicular low back pain, there was insufficient evidence from one poor-quality trial to determine effects of pregabalin versus amitriptyline (SOE: insufficient).
• For chronic nonradicular low back pain, one small trial found the addition of pregabalin 300 mg/day to transdermal fentanyl associated with substantially lower pain scores than transdermal buprenorphine alone at 3 weeks (difference ∼26 points on a 0 to 100 scale, p<0.05) but the estimate was very imprecise (SOE: insufficient).
• For chronic radicular pain, one trial found pregabalin (mean 2.1 mg/kg/day) plus celecoxib associated with lower pain scores than celecoxib alone (difference 11 points on a 0-100 scale, p=0.001) after 4 weeks and one trial found no effects of adding pregabalin (titrated to 300 mg/day) to tapentadol PR versus tapentadol PR alone on pain or the SF-12 after 8 weeks (SOE: insufficient).
• Two trials of gabapentin versus placebo reported no clear differences in risk of adverse events (SOE: low).
• Two trials of topiramate versus placebo reported inconsistent effects on risk of withdrawal due to adverse events; one of the trials found topiramate associated with higher risk of sedation and diarrhea (SOE: insufficient).
• Two trials of pregabalin versus placebo reported inconsistent effects on risk of withdrawal due to adverse events, somnolence, and dizziness; one of the trials used an enrichment/withdrawal design (SOE: insufficient).

Detailed Synthesis

The APS/ACP review²⁹ included four trials of antiseizure medications for low back pain.^156-159 Two trials (n=50 and 65)^{157, 159} evaluated gabapentin and two trials (n=29 and 96)^{156, 158} evaluated topiramate. All trials compared antiseizure medications versus placebo, with one trial¹⁵⁶ utilizing an “active” placebo (diphenhydramine). Three trials^{156, 157, 159} evaluated patients with radicular symptoms and one trial (of topiramate)¹⁵⁸ evaluated a mixed population of patients with radicular or nonradicular pain.

We identified seven trials of antiseizure medications for low back pain published subsequent to the APS/ACP review (Table 8; Appendix Tables E11, F11).^160-166 Sample sizes were 200 to 309 in the three largest trials^160-162 and ranged from 26 to 55 in the other four trials. Six trials^160-165 evaluated pregabalin and one trial¹⁶⁶ evaluated gabapentin. Two trials compared pregabalin versus placebo¹⁶⁰ or active placebo (diphenhydramine)¹⁶³ and one trial¹⁶⁶ compared gabapentin versus no gabapentin. Three trials compared pregabalin plus another medication (tapentadol,¹⁶¹ transdermal buprenorphine,¹⁶⁴ or celecoxib¹⁶⁵) versus the other medication without pregabalin and one trial¹⁶² compared pregabalin versus amitriptyline. The celecoxib trial also compared pregabalin alone versus celecoxib alone.¹⁶⁵ Five trials evaluated patients with radicular symptoms,^{160, 161, 163, 165, 166} with two trials^{163, 166} focusing on patients with spinal stenosis. One trial was restricted to patients with nonradicular back pain¹⁶⁴ and one trial enrolled a mixed population of radicular and nonradicular back pain.¹⁶²

Table 8

Characteristics and conclusions of included antiseizure trials.

All of the trials, including those in the APS/ACP review, evaluated patients with chronic symptoms. Dosing of antiseizure medications varied. One trial evaluated fixed-dose pregabalin 300 mg/day in combination with transdermal buprenorphine.¹⁶⁴ In the other pregabalin trials, doses were titrated, though titration protocols and maximum doses varied. Two trials^{156, 158} of topiramate titrated doses to 300 or 400 mg/day and three trials titrated gabapentin to a maximum dose that ranged from 1200 to 3600 mg/day.^{157, 159, 166} The duration of therapy ranged from 2 weeks to 4 months; outcomes were assessed at the end of or during therapy in all trials except for one,¹⁶¹ which evaluated patients 1-2 weeks after completing 8 weeks of therapy.

Three trials^{156, 163, 165} used a crossover design and the rest were parallel-group trials. Six trials^{158, 160, 161, 163-165} were rated fair quality and four^{156, 157, 159, 166} poor quality. Methodological shortcomings included inadequate description of randomization and allocation concealment methods and unclear blinding of outcome assessors. Additional shortcomings in the poor-quality trials included unblinded design or unclear blinding status, high attrition, and failure to perform intention-to-treat analysis. One trial of pregabalin used an enrichment/withdrawal design.¹⁶⁰ None of the crossover trials reported results of the first intervention period and two of the crossover trials^{163, 165} did not assess for carryover effects, though all employed a washout period between interventions.

We excluded one trial (n=113) of gabapentin (up to 3600 mg/day) versus placebo for chronic nonradicular pain only published as an abstract.¹⁶⁷

Key Points

• For acute nonradicular low back pain, two trials found no differences between a single intramuscular injection or a 5-day course of systemic corticosteroids versus placebo for pain or function (SOE: low for pain and function).
• For radicular low back pain (acute or unspecified duration) five trials consistently found no differences between systemic corticosteroids (administered a single bolus or as a short taper) versus placebo in pain or function; one trial found no effect on need for spine surgery (SOE: moderate for pain and function).
• For spinal stenosis, one trial found no differences through 12 weeks of followup between a 3-week course of prednisone versus placebo in pain intensity, the RDQ, or any SF-36 subscale (SOE: low for pain and function).
• Trials of systemic corticosteroids did not report serious adverse events, including hyperglycemia requiring medical treatment, but adverse events were not reported well in some trials (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included four trials of systemic corticosteroids.^168-171 Three trials (n=49 to 60)^{168, 170, 171} evaluated systemic corticosteroids in patients with radiculopathy and one trial (n=86)¹⁶⁹ evaluated patients with nonradicular back pain. For radiculopathy, the APS/ACP review concluded that there was consistent evidence that systemic corticosteroids were not associated with clinically significant benefits when given as a single large parenteral bolus or as a short oral or intramuscular taper.

We identified four trials^172-175 of systemic corticosteroids published subsequent to the APS/ACP review and one older trial¹⁷⁶ that was not included in the APS/ACP review (Table 9; Appendix Tables E12, F12). Three trials (n=27 to 78)^{173, 174, 176} evaluated patients with radicular pain, one trial (n=61) evaluated patients with spinal stenosis¹⁷⁵ and one trial (n=67)¹⁷² evaluated patients with nonradicular pain.

Table 9

Characteristics and conclusions of included corticosteroid trials.

All of the trials were placebo-controlled. Five trials^{168, 169, 172-174} evaluated patients with acute low back pain (including the two trials of nonradicular back pain)^{169, 172} and the other four^{170, 171, 175, 176} did not specify the duration of symptoms. Three trials were conducted in emergency department settings,^{169, 172, 173} one trial in an inpatient setting,¹⁶⁸ and in the other trials the clinical setting was not reported or mixed. The doses and mode of administration of corticosteroids varied. Three trials evaluated a single dose of parenteral (intravenous or intramuscular) methylprednisolone (150 to 500 mg, equivalent to 187.5 to 625 mg of prednisone).^{168, 169, 173} In the other trials, the duration of treatment ranged from 5 to 21 days. Three trials evaluated similar tapering courses of oral or intramuscular dexamethasone (64 mg for 1 day, 32 mg for 1 day, 24 mg for 1 day, 12 mg for 1 day, and 8 mg for 3 days [64 mg of dexamethasone equivalent to 400 mg of prednisone]).^{170, 171, 176} The other three trials evaluated different courses of oral prednisone (50 mg for 5 days,¹⁷² 60 mg for 3 days, 40 mg for 3 days, and 20 mg for 3 days,¹⁷⁴ or 1 mg/kg/day for 1 week, with a one-third dose reduction each week¹⁷⁵). The three single-dose trials evaluated patients at 10 days to 1 month after administration; in the other trials followup ranged from within 2 days after a 5 or 7 day course of therapy^{171, 172} to 1 to 4 years after a 1-week course of therapy.¹⁷⁰

Among the trials of systemic corticosteroids for radiculopathy, two trials^{173, 174} required a positive straight leg raise for inclusion and four others^{168, 170, 171, 176} required a positive straight leg raise or other signs of radiculopathy (e.g., sensory, motor, or reflex deficit). One of the latter trials also required imaging findings of a herniated disc that correlated with radicular symptoms.¹⁶⁸ The trial of patients with spinal stenosis required presence of neurogenic claudication symptoms and imaging findings of central stenosis.¹⁷⁵

Three trials^{168, 169, 173} were rated good quality, five trials fair quality,^{170-172, 175, 176} and one trial poor quality.¹⁷⁴ Methodological shortcomings in the fair-quality trials included inadequate description of allocation concealment, unclear blinding of outcomes assessors, and unclear compliance to interventions. The poor-quality trial allocated patients sequentially.¹⁷⁴

One trial (n=100) of dexamethasone versus placebo for nonradicular low back pain was excluded because it was published in German, but otherwise met inclusion criteria.¹⁷⁷

Key Points

• For acute low back pain, a systematic review found no differences between exercise therapy versus no exercise in pain (3 trials, WMD 0.59 at intermediate term on a 0 to 100 scale, 95% CI −11.51 to 12.69) or function (3 trials, WMD at short term −2.82, 95% CI −15.35 to 9.71 and WMD 2.47 at intermediate term, 95% CI −0.26 to 5.21). For subacute low back pain, there were also no differences in pain (5 trials, WMD 1.89 on a 100-point scale, 95% CI −1.13 to 4.91) or function (4 trials, WMD 1.07, 95% CI −3.18 to 5.32). Three subsequent trials for acute to subacute low back pain reported inconsistent effects of exercise versus usual care on pain and function (SOE: low for pain and function).
• For chronic low back pain, a systematic review found exercise associated with greater pain relief versus no exercise (19 trials, WMD 10 on a 0 to 100 scale, 95% CI 1.31 to 19.09), though the effect on function was small and not statistically significant (17 trials, WMD 3.00 on a 0 to 100 scale, 95% CI −0.53 to 6.48). Results from a more recent systematic review using more restrictive criteria and additional trials not included in the systematic reviews were generally consistent with these findings (SOE: moderate for pain and function).
• For chronic low back pain, a systematic review included two trials that found motor control exercise (MCE) associated with lower pain scores in the short term (WMD −12.48 on a 0 to 100 scale, 95% CI−19.04 to −5.93), intermediate term (WMD −10.18, 95% CI −16.64 to −3.72) and at long term (WMD −13.32 95% CI −19.75 to −6.90) versus a minimal intervention. MCE was also associated with better function at short term (3 trials WMD −9.00 on 0 to 100 scale, 95% CI −15.28 to −2.73), intermediate term (2 trials WMD −5.62, 95% CI−10.46 to −0.77) and long term (2 trials, WMD −6.64, 95% CI −11.72 to −1.57) (SOE: low for pain and function).
• For nonacute low back pain, a systematic review found no clear effects of exercise therapy versus usual care on likelihood of short- or intermediate-term (∼6 months) disability, but exercise was associated with lower likelihood of work disability at long-term (∼12 months) followup (10 comparisons in 8 trials, OR 0.66, 95% CI 0.48 to 0.92) (SOE: moderate for pain and function).
• For radicular low back pain, three trials not included in the systematic reviews found effects that favored exercise versus usual care or no exercise in pain and function, though effects were small (SOE: low for pain and function).
• For chronic low back pain, a systematic review found MCE associated with lower pain intensity at short term (6 trials, WMD −7.80 on 0 to 100 scale, 95% CI −10.95 to −4.65) and intermediate term (3 trials, WMD −6.06, 95% CI −10.94 to −1.18) versus general exercise, but effects were smaller and no longer statistically significant at long-term (4 trials, WMD −3.10, 95% CI −7.03 to 0.83). MCE was also associated with better function in the short term (6 trials, WMD −4.65 on 0 to 100 scale, 95% CI −6.20 to −3.11) and long term (3 trials, WMD −4.72, 95% CI −8.81 to −0.63). One of two subsequent trials found no effect on pain, though effects on function were consistent with the systematic review (SOE: low for pain and function).
• For comparisons involving other types of exercise techniques, there were no clear differences in >20 head-to-head trials of patients with acute or chronic low back pain (SOE: moderate).
• Harms were poorly reported in trials of exercise. When reported, harms were typically related to muscle soreness and increased pain, or no harms were reported; no serious harms were reported (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included six systematic reviews^178-184 with a total of 79 unique trials and one additional large, lower-quality trial.¹⁸⁵ The most comprehensive systematic review in the APS/ACP review found no differences between exercise therapy versus no exercise for acute low back pain in pain (3 trials, WMD 0.59 at intermediate term on a 0 to 100 scale, 95% CI −11.51 to 12.69) or function (3 trials, WMD at short term −2.82, 95% CI −15.35 to 9.71 and WMD 2.47 at intermediate term, 95% CI −0.26 to 5.21).¹⁷⁹ For subacute low back pain, there were also no differences in pain (5 trials, WMD 1.89 on a 100-point scale, 95% CI −1.13 to 4.91) or function (4 trials, WMD 1.07, 95% CI −3.18 to 5.32). For chronic low back pain, the APS/ACP review found good evidence that exercise is moderately superior to no exercise for pain relief (19 trials, WMD 10 on a 0 to 100 scale, 95% CI 1.31 to 19.09), though the effect on function was small and not statistically significant (17 trials, WMD 3.00 on a 0 to 100 scale, 95% CI −0.53 to 6.48). Results of the other reviews were generally consistent with these findings. Based on this evidence, the APS/ACP review concluded that there was fair evidence of no benefit for exercise versus no exercise for acute or subacute low back pain, and good evidence for moderate benefits of exercise versus no exercise for chronic low back pain.

We included three fair-quality systematic reviews of exercise for low back pain published subsequent to the APS/ACP review (Table 10; Appendix Tables E13, F13).^186-188 One focused on exercise for nonspecific chronic low back pain¹⁸⁷ one evaluated effects of exercise on work disability in patients with nonacute, nonspecific low back pain (duration >4 weeks),¹⁸⁸ and the third focused on effects of motor control exercise (MCE), which was not addressed in the APS/ACP review and not covered well in the other reviews.¹⁸⁶

Table 10

Summary of systematic reviews of nonpharmacological treatments for low back pain.

The first review focused on various types of exercise therapy for chronic low back pain and included 37 RCTs (n=3957).¹⁸⁷ Shortcomings of this review included limited description of included trial characteristics (including exercise treatments) failure to report statistical heterogeneity for pooled analyses, and lack of sensitivity or subgroup analyses. The systematic review included eight trials from a previous review¹⁷⁹ plus 29 additional trials; it excluded a number of trials in the previous review because it applied more strict criteria to define chronic low back pain (≥ 12 weeks), and only enrolled trials of patients with nonspecific low back pain. Exercise was compared against wait list/no treatment (8 RCTs), usual care (6 RCTs), back school or education (3 RCTs), and other forms of exercise therapy (11 RCTs). Exercise interventions varied and included general strengthening, stretching, or aerobic exercise; motor control and stabilization exercises; physiotherapy; multidisciplinary programs; and specific techniques such as the active trunk exercise protocol.¹⁸⁷ Comparisons of exercise versus other active interventions (behavioral treatment, passive modalities [TENS, laser, ultrasound, massage], spinal manipulations and psychotherapy), are discussed in the sections of this report addressing those interventions. Outcomes were assessed at the end of treatment, at short term (3 months), intermediate term (6 months), and long term (>6 months). Of the 27 trials providing data for the above comparisons,^189-215 11 (41%) were rated low risk of bias, based on meeting ≥6 of 11 Cochrane Back Review Group criteria.^{191-193, 197, 198, 201-203, 207, 209, 214, 216} Methodological flaws included failure to describe adequate randomization methods (26% of trials) or allocation concealment (48%), inadequate description of cointerventions(63%), unclear compliance with treatment (56%), failure to report intention-to-treat analysis (48%), and high or unreported attrition (33%). Given the nature of exercise interventions, blinding of patients and care providers was generally not possible; in addition, 67 percent of trials did not report blinding of outcome assessors.

The second review evaluated effects of exercise specifically on work disability in patients with nonacute (>4 weeks), nonspecific low back pain.¹⁸⁸ It included 23 trials (n=4138), 20 of which were included in meta analyses. Sample sizes ranged from 49 to 476 and duration of low back pain varied from 4 weeks to greater than 12 months. Exercise was compared against usual care (13 RCTs, n=3181) and other forms of exercise (11 RCTs). Exercise interventions varied and included stabilization, strengthening, stretching, and mobilization, though exercise regimens were most frequently mixed. About half of the exercise interventions were administered in the context of a cognitive behavioral approach. The majority of interventions (91%) were supervised exercise conducted in an outpatient setting (77%). Nine trials^{206, 217-224} (39%) were rated high quality and the remainder were rated low quality, based on risk of bias criteria by Juni et al. Methodological shortcomings included detection bias in 12 trials (52%), selection bias in 9 trials (39%), and attrition bias in 8 trials (35%).

The third review included 16 trials of MCE (sample size range 30 to 346, total n=1993).¹⁸⁶ MCE (also referred to as specific stabilization exercise) focuses on strengthening of deep muscles of the spine through a specific stabilization protocol, while reducing unwanted overactivity of other muscles.²²⁵ Methodological limitations of the review are that it did not report details regarding study quality (it reported an overall assessment only), did not report statistical heterogeneity in pooled analyses, and did not report harms. We addressed these issues through additional review and assessment of the primary studies. The review included seven trials of MCE versus various types of general exercise (including sling exercise, trunk strengthening, walking, cardiovascular and McKenzie exercises),^{192, 226-231} three trials of MCE versus a minimal intervention (no intervention, advice/education or placebo short-wave therapy and ultrasound),^{196, 232, 233} and four trials of MCE versus multimodal physical therapy (including ultrasound, electrotherapy, lumbar strengthening, passive physical therapy and general exercise).^{230, 234-236} Two trials evaluated MCE as part of a multimodal intervention versus other components of that intervention.^{237, 238} Eight trials included only chronic low back pain patients;^{192, 196, 226, 228-230, 232, 233} three trials also included patients with subacute low back pain, but mean duration of symptoms was 25 to 34 months.^{231, 234, 236} Three trials focused on patients with recurrent low back pain, with the duration of the current episode ranging from >6 weeks to >3 months.^{231, 235, 237} Most trials selected patients for inclusion on the basis of tests showing deficits in motor control, but the specific methods and criteria for inclusion varied. The duration of treatment ranged from 6 to10 weeks; six trials evaluated patients 10 to 28 months after the end of treatment. In two trials, MCE was administered through 20 treatment sessions (time frame not described), with followup for 180 days. The systematic review classified 10 trials as high quality, based on scoring ≥6 points on the 10 point PEDro scale^{192, 226, 227, 229, 230, 232-238} and six low quality.^{196, 227, 229, 231, 233, 234} Common methodological shortcomings included unclear or inappropriate randomization methods and unclear allocation concealment; patients and care providers generally could not be blinded. Some trials also reported discrepancies in baseline characteristics or differential attrition. Data were pooled for short term (≥6 weeks to <4 months) intermediate term (≥4 months to <8 months) and long term (≥8 months <15 months).

We identified 17 trials with sample sizes >100 of exercise for low back pain that were not included in the systematic reviews (Table 11; Appendix Tables E14, F14).^239-255 Three trials evaluated exercise versus no exercise or usual care for acute to subacute low back pain,^{244, 254, 255} two trials compared different types of exercise for patients primarily with subacute low back pain,^{243, 246} four trials compared exercise versus no exercise or usual care for chronic low back pain,^{250-252, 256} four trials compared different types of exercise for chronic low back pain,^{241, 249, 251, 253} and four trials evaluated exercise versus various other interventions for radicular low back pain.^{239, 240, 242, 248} Exercise techniques varied but included general exercise, strengthening, the McKenzie method, exercise based on a treatment-based classification (TBC) system, the Alexander technique, periodized musculoskeletal rehabilitation, walking, MCE and others. Four trials were rated good quality,^{241, 242, 250, 252} seven trials fair quality,^{239, 240, 244, 245, 249, 253-255} and five poor quality.^{243, 246, 248, 251, 256} Methodological shortcomings included inadequate allocation concealment, failure to clearly described cointerventions, and failure to report compliance to treatment. Patients and people administering exercise could not be effectively blinded given the nature of the interventions.

Table 11

Characteristics and conclusions of included exercise trials.

We also identified 28 trials not included in the systematic reviews that evaluated exercise therapy for subacute to chronic low back pain, but enrolled fewer than 100 patients. Eighteen trials compared exercise versus no exercise or usual care^{197, 200, 257-272} and 15 compared different forms of exercise.^{197, 200, 258, 261, 271, 273-282} Given the number of larger trials on exercise, we did not abstract these studies in detail.

Key Points

• For chronic low back pain, a systematic review included seven trials that found Pilates associated with small (mean difference −1.6 to −4.1 points) or no clear effects on pain at the end of treatment versus usual care plus physical activity and no clear effects on function (SOE: low for pain and function).
• For chronic low back pain, three trials found no clear differences between Pilates versus other types of exercises in pain or function (SOE: low for pain and function).

Detailed Synthesis

The previous APS/ACP review did not specifically evaluate Pilates. A systematic review on exercise therapies included in the APS/ACP review did not include any studies of Pilates.

A fair-quality systematic review published subsequent to the APS/ACP review²⁸⁶ included seven trials of Pilates versus usual care (sample size range17 to 86, total n=301)^{195, 287-292} and four trials of Pilates versus other exercise techniques (sample size range 12 to 83, total n=199) (Table 10; Appendix Tables E15; F15).^293-296 The trials exclusively or primarily (∼75%)^{292, 293} enrolled patients with chronic low back pain. Pilates interventions varied but generally included one or three supervised mat small group classes per week plus home sessions; some included specific Pilates equipment. Usual care was generally less well described, but typically involved no specific treatment apart from medications and no restriction from regular physical activity. One study allowed both groups to continue physical therapy and regular exercise²⁸⁹ and another provided an educational booklet on low back pain.²⁹⁰ Exercise techniques in trials of Pilates versus other exercise methods included supervised stationary cycling, traditional lumbar stabilization exercises, McKenzie exercises and a generalized exercise regimen that included aerobics, stretching and strengthening. The duration of interventions in the trials ranged from four to 12 weeks. Three trials followed participants 16 to 18 weeks beyond the end of the active intervention. Most trials were conducted in Brazil, Australia and the United Kingdom and three trials were published as dissertations.^{289, 293, 297} Based on the 16-item McMaster Critical Review Form for Quantitative Studies, review authors classified four trials excellent (15 or 16 out of 16 points),^{289, 290, 294, 296} four very good (13-14 points),^{195, 287, 291, 292} one fair (9-10 points)²⁹³ and four poor (0-8 points).^{288, 295-298} Methodological shortcomings included inability to blind patients (most trials blinded outcomes assessors) and high attrition in trials with longer followup.^{294, 296}

Key Points

• For chronic low back pain, two trials found tai chi associated with improved pain-related outcomes versus wait list or no tai chi (mean differences 0.9 and 1.3 on a 0 to 10 scale); one trial also found tai chi associated with better function (mean difference 2.6 on the RDQ, 95% CI 1.1 to 3.7) (SOE: low for pain and function).
• For chronic low back pain, one trial found tai chi associated with lower pain intensity versus backward walking or jogging through 6 months (mean differences −0.7 and −0.8), but there were no differences versus swimming (SOE: low).
• One trial of tai chi reported a small temporary increase in back pain symptoms and one trial reported no harms (SOE: low).

Detailed Synthesis

Tai chi was not specifically evaluated in the APS/ACP review. We identified two trials of tai chi versus no treatment for chronic low back pain with no treatment (2 RCT, n=480) (Table 12; Appendix Tables E16, F16;^{285, 299} one of the trials also evaluated tai chi versus other exercise interventions including backward walking, jogging and swimming.²⁹⁹ Tai chi sessions were eighteen 40-minute sessions over 10 weeks in one trial²⁸⁵ and 45-minute sessions 5 days a week for 6 months in the other.²⁹⁹ Both trials were rated fair quality. One trial did not adequately report allocation concealment and attrition²⁹⁹ and adherence was unclear in both trials. The nature of the intervention precluded blinding of participants and people administering the interventions, but both trials reported blinding of outcomes assessors.

Table 12

Characteristics and conclusions of included tai chi trials.

Key Points

• For chronic low back pain, one trial found Iyengar yoga associated with lower pain scores (24 vs. 37 on a 0-100 VAS, p<0.001) and better function (18 vs. 21 on the 0 to 100 ODI, p<0.01, on a 0 to 100 scale) versus usual care at 24 weeks (SOE: low for pain and function).
• For chronic low back pain, a systematic review found yoga associated with lower pain intensity and better function versus exercise in most trials, though effects were small and differences were not always snot statistically significant (5 trials) (SOE: low for pain and function).
• For chronic low back pain, yoga was associated with lower short-term pain intensity versus education (5 trials, SMD −0.45,- 95% CI −0.63 to −0.26; I²=0%), but effects were smaller and not statistically significant at longer-term followup (4 trials, SMD −0.28, 95% CI−0.58 to −0.02, I²=47%); yoga was also associated with better function at short-term (5 trials, SMD 0.45, 95% CI −0.65 to −0.25; I²=8%) and long-term followup (4 trials, SMD 0.39, 95% CI −0.66 to −0.11; I²=40%) (SOE: moderate for pain and function).
• Reporting of harms was suboptimal, but adverse events when reported were almost all classified as mild to moderate (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included three trials (n=22 to 101) of yoga for chronic low back pain.^{194, 207, 213} One trial evaluated Viniyoga¹⁹⁴ and two trials Iyengar yoga;^{207, 213} comparator interventions were exercise or self-care. The APS/ACP review concluded that there was fair evidence that Viniyoga is moderately effective for chronic low back pain, with insufficient evidence to judge the effectiveness of other yoga styles, or the effectiveness of yoga for acute low back pain.

A good-quality systematic review³⁰⁰ published subsequent to the APS/ACP review included 10 trials,^{194, 207, 210, 213, 301-306} including the 3 trials described above (Table 10; Appendix Tables E17, F17). All trials enrolled patients with chronic (>3-month duration) low back pain, except for one small trial (n=12) which enrolled patients with back pain for >3 weeks. Sample sizes ranged from 12 to 313 (total sample=1,056). All yoga interventions included specific asanas (poses), pranayama (breathing), and relaxation, and many included meditation or mental focus practices. The most common specific yoga styles evaluated were Iyengar (5 trials) and Viniyoga (2 trials). Most trials evaluated yoga classes lasting 75 minutes once weekly with recommended home practice for 30 minutes 5 to 7 days per week, though one trial²¹⁰ evaluated all-day sessions over a 1-week period. Trials generally reported starting out with simple or restorative yoga poses and progressing to more challenging poses. The duration of active intervention ranged from 1 to 24 weeks. Outcomes were assessed at the end of therapy in all trials; five trials also assessed outcomes 14 to 52 weeks after the end of therapy. Yoga was compared versus usual care (2 trials), education (7 trials), and supervised exercise therapy (3 trials). Exercise therapy interventions varied, but included stretching, strengthening, and aerobic exercise. Two trials were conducted in India,^{210, 302} one in the United Kingdom,³⁰¹ and the remainder in the United States. Two trials^{213, 302} were rated high risk of bias (based on meeting fewer than 6 of 10 Cochrane Back Review Group criteria) and the remainder were rated low risk of bias; methodological shortcomings included inadequate reporting of randomization and allocation concealment methods and high attrition. Blinding of patients and caregivers was generally not possible, though 8 of the 10 trials reported blinding of outcome assessors.

We identified two additional trials^{307, 308} not included in the systematic review of yoga for chronic low back pain (Table 13; Appendix Tables E18; F18).^{307, 308} One Indian trial (n=60) compared a 60-minute class of Iyengar Yoga per week for 4 weeks (plus home practice) versus exercises (primarily strengthening exercises)³⁰⁷ and a US trial (n=95) compared once versus twice weekly 75-minute hatha yoga classes over 12 weeks.³⁰⁸ Both trials were rated fair quality; methodological shortcomings included unclear allocation concealment methods and unblinded design.

Table 13

Characteristics and conclusions of included yoga trials.

Key Points

• For chronic low back pain, a systematic review found progressive relaxation superior to wait list control for post-treatment pain intensity (3 trials, mean difference −19.77 on 0 to 100 VAS, 95% CI -34 to −5.20, I²=57%) and functional status (3 trials, standardized mean difference −0.88, 95% CI −1.36 to −0.39, I²=0%) (SOE: low for pain and function).
• For chronic low back pain, a systematic review found electromyography (EMG) biofeedback associated with lower pain intensity at the end of treatment (3 trials, SMD −0.80, 95% CI −1.32 to −0.28, I²=0%), with no clear effect on function (3 trials) (SOE: low for pain and function).
• For chronic low back pain, a systematic review found operant therapy associated with lower pain intensity at the end of treatment (3 trials, standardized mean difference −0.43, 95% CI −0.75 to −0.1, I²=0%), with no clear effect on function (2 trials) (SOE: low for pain and function).
• For chronic low back pain, there was insufficient evidence from two trials to determine effects of cognitive therapy versus wait list control, due to inconsistency and imprecision (SOE: insufficient).
• For chronic low back pain, a systematic review found cognitive-behavioral and other combined psychological therapy associated with greater improvements in post-treatment pain intensity compared with wait list control (5 trials, SMD −0.60, 95% CI −0.97 to −0.22, I²=40%), but effects on function were smaller and not statistically significant (4 trials, SMD −0.37, 95% CI −0.87 to 0.13, I²=50%) (SOE: low for pain and function).
• For chronic low back pain, a systematic review found no clear differences between psychological therapies versus exercise therapy in pain intensity (2 trials) or between psychological therapies plus physiotherapy versus physiotherapy alone (6 trials) in pain or function, though one small subsequent trial found combination therapy associated with greater improvements in pain and function immediately after treatment (SOE: low for pain and function).
• Ten trials found no clear differences between different psychological therapies in pain or function (SOE: moderate for pain and function).
• Harms were not well-reported, but no trial included reported any adverse events associated with psychological therapies (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included two high-quality systematic reviews on psychological therapies for chronic low back pain.^{311, 312} One review included 22 trials (6 assessed as higher quality)³¹¹ and the other included 21 trials (7 assessed as higher quality).³¹² Together, the two reviews included a total of 35 unique studies. Based on the systematic reviews, the APS/ACP review concluded that there was good evidence that versus no psychological therapy or wait-list control, cognitive-behavioral therapy is associated with moderate benefits, good evidence that operant therapy is associated with no effect, fair evidence that progressive relaxation is associated with substantial net benefits, and insufficient evidence to determine effects of biofeedback. Neither systematic review found any differences between one type of behavioral intervention versus another.

An updated version of one of the reviews³¹² included in the APS/ACP review has been published (Table 10; Appendix Tables E19, F19).³¹³ It included 28 trials relevant to this report (total n=3090, sample sizes ranged from 18 to 409). Compared with the previous version, the updated review included seven additional trials and excluded three previously included trials. The review focused on psychological therapies conducted in an office or group setting, broadly classified into respondent (10 trials), operant (7 trials), cognitive (4 trials), and cognitive-behavioral (7 trials) treatments as well as combinations thereof (8 trials).³¹³ Operant therapies refer to behavioral therapies that encourage healthy behaviors such as exercise and participation in usual activities, and that do not reinforce patient pain behaviors. Cognitive therapies help patients to identify and challenge maladaptive thoughts that contribute to disability and distress. Respondent therapy includes techniques such as relaxation or biofeedback, and is based on the premise that the physiological response to pain is linked to muscle tension in a negative feedback loop, and that this cycle can be interrupted by reducing muscle tension. Twelve trials compared psychological therapies versus wait list control, seven trials compared psychological therapies versus other interventions, and 10 trials compared one psychological therapy versus another. Several trials evaluated more than one type of psychological therapy. The duration and intensity of treatments were inconsistently described; when reported they varied from 35- to 120-minute sessions over 3 to 10 weeks; one trial evaluated daily 8-hour treatments over 5 weeks. Outcomes were assessed during or at the end of treatment in 25 trials and at 3 to 24 months after treatment in 21 trials.

Thirteen trials were classified as being at low risk of bias (based on meeting at least 6 of the 12 Cochrane Back Review Group criteria). Common methodological shortcomings included inadequate description of randomization and allocation concealment methods, high attrition, and dissimilar cointerventions among groups. The majority of trials used an unblinded design.³¹³

We identified five additional trials of psychological therapies for chronic low back pain not included in the systematic review (Table 14; Appendix Tables E20; F20).^314-318 One trial evaluated psychological therapy versus wait list control³¹⁷ and three trials (across four publications) psychological therapies plus another noninvasive intervention versus the other intervention alone.^{314-316, 318} All of the trials were rated fair quality. In general, neither patients nor care providers could be blinded, compliance to treatment was low or unreported, and some trials did not report allocation concealment methodology or use of intention-to-treat analysis.^315-318

Table 14

Characteristics and conclusions of included psychological therapy trials.

Key Points

• For chronic low back pain, a systematic review found multidisciplinary rehabilitation, versus usual care, associated with lower short-term pain intensity (9 trials, standardized mean difference −0.55, 95% CI −0.83 to −0.28, I²=72%; or ∼1.4-point mean difference on a 0- to 10-point numerical rating scale) and disability (9 trials, standardized mean difference −0.41, 95% CI −0.62 to −0.19, I²=58%; or ∼2.5-point mean difference on the RDQ); effects on long-term pain intensity and disability also favored multidisciplinary rehabilitation, but were smaller (7 trials, standard mean difference −0.21, 95% CI −0.37 to −0.04, I²=25% and 6 trials, standardized mean difference −0.23, 95% CI −0.40 to −0.06, I²=19%, respectively), with no difference in likelihood of return to work (7 trials, OR 1.04, 95% CI 0.73 to 1.47, I²=31%) (SOE: moderate for pain and function).
• For chronic low back pain a systematic review found multidisciplinary rehabilitation, versus no multidisciplinary rehabilitation, associated with lower short-term pain intensity (3 trials, standardized mean difference −0.73, 95% CI −1.22 to −0.24, I²=64%, or ∼1.7-point mean difference on a 0 to 10 numerical rating scale) and disability (3 trials, pooled standardized mean difference −0.49, 95% CI −0.76 to −0.22, I²=0%, or ∼2.9-point mean difference on the RDQ); there was insufficient evidence to assess effects on long-term outcomes (SOE: low for pain and function).
• For chronic low back pain, a systematic review found multidisciplinary rehabilitation, versus nonmultidisciplinary physical therapy, associated with lower short-term pain intensity (12 trials, standardized mean difference −0.30, 95% CI −0.54 to −0.06, I²=80%, or an approximate 0.6-point mean difference on a 0- to 10-point numerical rating scale) and disability (13 trials, standardized mean difference −0.39, 95% CI −0.68 to −0.10, I²=88%, or an approximate 1.2-point mean difference on the RDQ); multidisciplinary rehabilitation was also associated with lower long-term pain intensity (9 trials, standardized mean difference −0.51, 95% CI −1.04 to 0.01, I²=92%) and function (10 trials, standardized mean difference −0.68, 95% CI −1.19 to −0.16, I²=94%) and greater likelihood for return to work (8 trials, OR 1.87, 95% CI 1.39 to 2.53, I²=0%) (SOE: moderate).
• No study evaluated the effectiveness of multidisciplinary rehabilitation for acute low back pain or for radicular low back pain.
• Harms were poorly reported in trials of multidisciplinary rehabilitation, though no serious harms were reported (SOE: low).

Detailed Synthesis

Multidisciplinary rehabilitation, also known as interdisciplinary rehabilitation, refers to a coordinated program with both physical and biopsychosocial treatment components (at minimum) and is provided by professionals from at least two different specialties (e.g., physical therapists, occupational therapists, psychologists, physicians, and/or complementary and alternative medicine providers). The previous APS/ACP review²⁹ identified three systematic reviews of multidisciplinary rehabilitation for chronic low back pain (>3 months duration) and one trial of multidisciplinary rehabilitation for subacute (>4 weeks and <3 months duration) low back pain.^344-347 The systematic reviews were all rated high quality and included 20 unique trials. Based on the systematic reviews, the APS/ACP review concluded that there was good evidence that multidisciplinary rehabilitation interventions for chronic low back pain are moderately more effective than usual care or no multidisciplinary intervention at reducing pain and improving function, including return to work.^344-347

We identified a subsequent good-quality systematic review of multidisciplinary rehabilitation that included 41 trials of multidisciplinary biopsychosocial rehabilitation (MBR) for chronic (>12 weeks) mechanical or nonspecific low back pain (Table 10; Appendix Tables E21; F21).³⁴⁸ Thirty-one of the trials were published after the APS/ACP review. The trials in the systematic review enrolled a total of 6,858 subjects (sample size range 20 to 542) from Europe, Iran, North America, and Australia. Sixteen trials compared MBR versus usual care; 4 trials MBR versus wait list controls; 19 trials MBR versus physical treatments (exercise plus other modalities like back school, massage, traction, and stretching); and 12 trials compared different multidisciplinary rehabilitation interventions versus one other. Trials of MBR versus surgery were included in the systematic review but outside the scope of this report. Fifteen of the MBR interventions were categorized as high-level interventions (>100 hours total and delivered on a daily basis), 15 involved low-level interventions (<30 hours and nondaily), and 11 interventions did not meet criteria for either high- or low-level interventions. Primary outcomes of pain, disability, and work were organized into short-term outcomes (<3 months), medium term-outcomes (3-12 months), and long-term outcomes (≥12 months). All of the studies had methodological shortcomings, but 13 trials were assessed by the review as low risk of bias, based on meeting ≥6 of 12 Cochrane Back Review criteria. No trial blinded providers or participants; other methodological shortcomings included failure to describe adequate randomization methods (12 trials) and failure to report intention to treat analysis (25 trials).

We identified three additional trials of multidisciplinary rehabilitation that were not included in the systematic review (Table 15; Appendix Tables E22, F22).^349-351 Two trials evaluated multidisciplinary rehabilitation for subacute (<12 weeks duration) low back pain.^{349, 350} One good-quality trial (n=20) compared a low-intensity multidisciplinary program consisting of physician evaluation, acupuncture, chiropractic care, massage, occupational therapy, physical therapy, nutrition counseling, and as-needed psychiatric and rheumatology consults versus usual care.³⁴⁹ A fair-quality trial (n=70) compared a high-intensity (>100 total hours) multidisciplinary rehabilitation program including physician evaluation, physical therapy, biofeedback/pain management, group didactic sessions, case management/occupational therapy sessions, and interdisciplinary team conference for patients at high risk for chronic disabling low back pain versus usual care.³⁵⁰ A third, good-quality trial (n=20) evaluated multidisciplinary rehabilitation (including exercise and cognitive-behavioral therapy) versus usual care (including passive spinal mobilization and exercise) for chronic low back pain.³⁵¹

Table 15

Characteristics and conclusions of included multidisciplinary rehabilitation trials.

Key Points

• For acute low back pain, a systematic review found acupuncture associated with lower pain intensity versus sham acupuncture using nonpenetrating needles (2 trials, mean difference 9.38 on a 0 to 100 VAS, 95% CI 1.76 to 17.0, I²=27%); three other trials reported effects consistent with these findings. One trial of sham acupuncture using penetrating needles to nonacupuncture points found no effect on pain. These were no clear effects on function in 5 trials (SOE: low for pain and function).
• For chronic low back pain, a systematic review found acupuncture associated with lower pain intensity versus sham acupuncture (superficial needling at acupuncture or nonacupuncture points, or nonpenetrating pressure at acupuncture points) immediately at the end of treatment (4 trials, WMD −16.76, 95% CI −33.3 to − 0.19, I²=90%) and at up to 12 weeks (3 trials, WMD −9.55, 95% CI −16.5 to −2.58, I²=40%), but there were no differences in function. Four additional trials reported results consistent with these findings (SOE: moderate for pain and function).
• For chronic low back pain, a systematic review found acupuncture associated with lower pain intensity (4 trials, SMD −0.72, 95% CI −0.94 to −0.49, I²=51%) and better function (3 trials, SMD −0.94, 95% CI −1.41 to −0.47, I²=78%) immediately after treatment versus no acupuncture. Mean effects on pain ranged from 7 to 24 points on a 0- to 100-point scale; for function one trial reported a difference of 8 points on a 0- to 100-point scale and the other two trials; two trials showed small or no clear differences at longer-term followup (SOE: moderate for pain and function).
• For acute low back pain, a systematic review found acupuncture associated with slightly greater likelihood of overall improvement versus NSAIDs at the end of treatment (5 trials, RR 1.11, 95% CI 1.06 to 1.16, I²=0%) (SOE: low).
• For chronic low back pain, a systematic review found acupuncture associated with better pain relief (3 trials, WMD −10.56 on a 0 to 100 scale, 95% CI −20.34 to −0.78, I²=0%) and improvement in function (3 trials, SMD −0.36, 95% CI −0.67 to −0.04, I²=7%) immediately postintervention (SOE: low).
• Harms of acupuncture were poorly reported in the trials, though no serious adverse events were reported (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included three systematic reviews^361-363 with a total of 51 unique trials of acupuncture. Four trials in the systematic reviews evaluated acupuncture for acute low back pain and the remainder evaluated acupuncture for chronic low back pain. Based on the evidence in the systematic reviews, the APS/ACP review found insufficient (poor) evidence to determine effects of acupuncture for acute low back pain and fair evidence of moderate effects of acupuncture versus sham or no acupuncture for short-term pain relief in patients with chronic low back pain, though some inconsistency was noted in trials of acupuncture versus sham acupuncture, with some trials findings no effects.

We identified two fair-quality recent systematic reviews of acupuncture for low back pain; one evaluated acupuncture for acute or subacute low back pain³⁶⁴ and the other evaluated acupuncture for chronic low back pain (Table 10, Appendix Tables E23, F23).³⁶⁵

The systematic review³⁶⁴ on acupuncture for acute or subacute low back pain 9<12 weeks in duration) included 11 trials (9 not included in the APS/ACP review).^366-374 Three trials evaluated acupuncture versus sham acupuncture (total n=148), 7 trials evaluated acupuncture versus medications including NSAIDS, muscle relaxants and analgesics (total n=966). and 1 trial compared acupuncture plus medication versus the medication alone (n=49). The acupuncture interventions ranged from a single session^{371, 375} to up to 12 sessions over a 4- to 6-week period. Outcomes were assessed immediately at the end of treatment in all trials; longer-term outcomes were assessed at 1 to 6 months in three trials.^{374, 376, 377} Five trials were rated low risk of bias,^{366, 371, 374-376} based on meeting at least 6 of 12 2009 Cochrane Back Group criteria. Methodological shortcomings in the 6 high-risk of bias trials included inadequate description of randomization and allocation concealment methods, unblinded design, and unclear similarity of the groups at baseline. Three sham-controlled trials had blinding of patients, providers and outcomes assessors.^{371, 374, 375}

The systematic review³⁶⁵ on acupuncture for chronic low back pain included 32 trials (9 not included in the APS/ACP review).^378-385 All of the trials evaluated patients with chronic low back pain for >12 weeks, with the exception of one trial that included people with subacute (>6 weeks) to chronic low back pain (up to 52 weeks).³⁸⁶ In addition to standard acupuncture needles applied to the body, other acupuncture techniques evaluated in the trials included electroacupuncture and auricular acupuncture. Seven trials evaluated acupuncture versus sham procedures (total sample=638 participants), 3 trials evaluated acupuncture versus medications (total sample=75 participants), and the other 22 trials compared acupuncture versus no acupuncture, usual care, TENS, exercise, inactive treatment, or another active treatment. The number of sessions ranged from 1 to 20, the duration of treatment ranged from 1 day (single treatment) to 12 weeks, and duration of followup ranged from immediately following treatment through up to 48 months. Seven of the trials^{378-381, 383, 387, 388} were rated low risk of bias (based on meeting all 2009 Cochrane Back Review Group criteria). Methodological shortcomings in the other trials included inadequate description of randomization and allocation concealment techniques, unblinded design, and unclear similarity of groups at baseline.

We identified three additional good-quality trials of acupuncture for acute^{389, 390} or chronic³⁹¹ low back pain (Table 16, Appendix Tables E24, F24). One trial (n=80) compared five 30-minute sessions scalp acupuncture plus diclofenac versus sham scalp acupuncture plus diclofenac for acute low back pain; outcomes were assessed at 28 days.³⁸⁹ Another trial (n=270), randomized patients with acute low back pain to one of four treatment groups: true acupuncture, sham acupuncture (needles inserted at nonacupuncture points), placebo acupuncture (momentary pressure with semiblunted needle applied to back) or no acupuncture.³⁹⁰ Treatments were administered in five 20-minute sessions over 2 weeks, with outcomes assessed through 48 weeks. The third trial (n=130) evaluated acupuncture versus sham acupuncture for chronic low back pain.³⁹¹ Patients received 12 acupuncture or sham acupuncture sessions over a 6 week time period and were followed for up to 6 months.

Table 16

Characteristics and conclusions of included acupuncture trials.

We also identified one fair-quality trial (n=236) of acupuncture performed at back pain specific acupoints or standard acupuncture performed at nonspecific acupoints (n=82) versus usual care.³⁹² Methodological limitations included unclear allocation concealment and lack of blinding of patients and providers. Patients in the acupuncture groups received 14 daily treatments and outcomes were assessed through 24 weeks. A poor-quality trial (n=143) (due to unclear randomization and allocation concealment methods, no primary outcome identified and unclear blinding) compared the addition of daily acupuncture to an intensive inpatient 21-day rehabilitation for chronic low back pain and measured outcomes at 3 months after the end of treatment.³⁹³

Key Points

• For subacute low back pain, a systematic review included two trials that found massage associated with greater short-term (1 week) improvement in pain (SMD −0.92, 95% CI −1.35 to −0.48) and function (SMD −1.76, 95% CI −3.19 to −0.32) versus sham therapy, but there was no difference in pain or function at 5 weeks in one trial (SOE: low for pain and function).
• For chronic low back pain, one trial found no difference between foot reflexology versus usual care in pain or function, and one trial found structural or relaxation massage associated with better function (mean 2.5 to 2.9 points on the RDQ) versus usual care at 10 weeks; effects were less pronounced at 52 weeks (SOE: low for pain and function).
• For subacute to chronic low back pain, a systematic review found massage associated with better effects on short-term pain in 7 of 9 trials (mean differences −0.6 to −0.94 points on a 0 to 10 scale) and better effects on short-term function in 3 of 4 trials (SOE: moderate for pain and function).
• For subacute to chronic low back pain, a systematic review included 5 trials that generally found massage plus another intervention superior to the other intervention without massage for short-term pain, with effects somewhat stronger in trials in which massage was combined with exercise; few differences were observed for function or long-term pain. Two subsequent trials of massage plus exercise reported findings generally consistent with these findings (SOE: low).
• Comparisons of difference massage techniques were too heterogeneous and effects were too small from six trials to determine effects on pain and function (SOE: insufficient).
• Harms were not well-reported in trials of massage, though no serious adverse events were reported; two trials reported soreness during or shortly after the treatment (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included two good-quality systematic reviews with a total of eight unique trials of massage.^{403, 404} Five of the trials were rated higher quality. Based on the systematic reviews, the APS/ACP review concluded that there was fair evidence of moderate net benefits of massage for chronic or subacute low back pain.

One of the systematic reviews⁴⁰⁴ has been updated to include 13 trials (total n=1596, range 39 to 262) (Table 10; Appendix Tables E25, F25).⁴⁰⁵ The trials evaluated massage for acute (1 RCT), subacute (4 RCTs) and chronic low back pain (8 RCTs). Massage techniques were variable, and included traditional Thai massage, Swedish massage, relaxation massage methods, acupuncture massage, muscle energy technique, roptrotherapy, acupressure, foot reflexology, or combined techniques. Two trials compared massage versus sham/placebo massage, nine trials of massage versus other treatments (manipulation, exercise, relaxation, acupuncture, physiotherapy and self-care education), five trials of massage versus other interventions versus the other interventions alone, and two trials compared different massage techniques. The number of sessions, duration of sessions, and duration of treatment varied. Two of the studies included were single intervention trials; in the remainder the duration of treatment ranged from three to 10 weeks. The duration of followup ranged from immediately following treatment to 52 weeks post randomization. Six trials were rated low risk of bias (based on meeting ≥6 of 11 Cochrane Back Review group criteria). Methodological shortcomings included lack of blinding of patients and assessors and inadequate or unclearly described allocation concealment methods.

We identified eight additional trials of massage not included in the systematic review.^406-413 One good-quality trial (n=401) evaluated two different types of massage (structural or relaxation) versus usual care for chronic low back pain (Table 17; Appendix Tables E26, F26).⁴⁰⁶ Patients randomized to massage received 10 weekly treatments including up to 3 home exercises from a predefined list of seven exercises, six of which were common to both treatments, as well as stretching. The relaxation massage group was also given 2.5-minute home relaxation exercises. A fair-quality trial (n=45) compared acupressure massage versus sham laser or no treatment.⁴¹² Three smaller trials (n=26 to 140) not compared different massage techniques with one another. One trial was rated as good quality,⁴⁰⁷ one fair,⁴⁰⁹ and one poor quality.⁴⁰⁸ Two fair-quality trials compared massage plus an exercise intervention to the exercise intervention alone.^{411, 413} One of these trials trial (n=80)⁴¹¹ evaluated myofascial release for subacute to chronic low back pain and the other⁴¹³ evaluated Chinese massage for back pain of mixed duration. The eighth, fair-quality trial (n=32) compared massage plus traction versus traction alone in patients with chronic low back pain.⁴¹⁰ Methodological shortcomings in the fair- and poor-quality trials included unclear randomization and allocation concealment methods, baseline group differences, and inadequate or unclear blinding.

Table 17

Characteristics and conclusions of included massage trials.

Key Points

• For acute low back pain, two trials (one included in a systematic review) found spinal manipulation associated with better effects on function versus sham manipulation (statistically significant in one trial); in one trial effects on pain favored manipulation but were small and not statistically significant (mean difference −0.50, 95% CI −1.39 to 0.39) (SOE: low for function, insufficient for pain).
• For chronic low back pain, a systematic review found spinal manipulation associated with small, statistically nonsignificant effects versus sham manipulation on pain at 1 month (3 trials, WMD −3.24, 95% CI −13.62 to 7.15 on a 0 to 100 scale, I²=53%); one trial reported similar results for function (SMD −0.45, 95% CI −0.97 to 0.06); one trial not included in the systematic review reported generally consistent results (SOE: low for pain, insufficient for function).
• For acute low back pain, a systematic review found no differences between spinal manipulation versus and inert treatment in pain relief at 1 week (3 trials, WMD 0.14 on a 0 to 10 scale, 95% CI −0.69 to 0.96, I²=27%), though one trial found SMT associated with better longer-term pain relief (MD −1.20 at 3 months, 95% CI 2.11 to −0.29); there were no differences in function at 1 week (2 trials, SMD −0.08, 95% CI −0.37 to 0.21, I²=0%) or at 3 months (1 trial, SMD −0.28, 95% CI −0.59 to 0.02) (SOE: low for pain and function).
• For chronic low back pain, one high-quality trial found spinal manipulation associated with greater improvement in the “main complaint” versus an inert treatment (mean difference 0.9 on a 0 to 10 scale, 95% CI 0.1 to 1.7); results from three low risk of bias trials and three additional trials not included in the systematic review were somewhat inconsistent, though some trials reported effects that favored manipulation (SOE: low).
• For acute low back pain, a systematic review found no difference between spinal manipulation versus other active interventions in pain relief at 1 week (3 trials, WMD 0.06 on a 0 to 10 scale, 95% CI −0.53 to 0.65, I²=0%), 1 month (3 trials, WMD −0.15, 95% CI −0.49 to 0.18, I²=0%), 3 to 6 months (2 trials, WMD−0.20, 95%CI −1.13 to 0.73, I²=81%), or 1 year (1 trial, MD 0.40, 95% CI −0.08 to 0.88). Findings were similar for function, with no differences observed at any time point. A subsequent trial of patients with acute or subacute low back pain found spinal manipulation associated with moderate effects versus usual care on pain and small effects on function at short-term followup, but effects were smaller and no longer statistically significant at 3 and 6 months (SOE: moderate for pain and function).
• For chronic low back pain, a systematic review found spinal manipulation associated with better short-term pain relief versus other active interventions at 1 month (10 comparisons from 6 trials, WMD −2.76 on a 0 to 100 scale, 95% CI −5.19 to −0.32, I²=27%) and 6 months (7 comparisons from 4 trials, WMD −3.07, 95% CI −5.42 to −0.71, I²=0%), though the magnitude of effects was below the small/slight threshold. There was no difference at 12 months (3 trials, WMD −0.76, 95% CI −3.19 to 1.66, I²=0%). Manipulation was also associated with greater function improvement in function versus other active interventions at 1 month (10 comparisons from 6 trials, SMD −0.17, 95% CI −0.29 to −0.06, I²=3%); effects were smaller and no longer statistically significant at 6 and 12 months. Three trials not included in the systematic reviews reported results consistent with these findings (SOE: moderate for pain and function).
• For acute low back pain, four trials in a systematic review found spinal manipulation plus either exercise or advice associated with greater improvement in function at 1 week (SMD −0.41, 95% CI −0.73 to −0.10, I²=18%) versus exercise or advice alone, but there were no differences at 1 month (3 trials, SMD −0.09, 95% CI −0.39 to 0.21, I²=37% ) or 3 months (2 trials, SMD −0.22, 95% CI −0.61 to 0.16, I²=41%) (SOE: low).
• For chronic low back pain, a systematic review found spinal manipulation plus another active treatment associated with greater pain relief at 1 month (3 trials, WMD −5.88 on a 0 to 100 scale, 95% CI −10.85 to −0.90, I²=0%), 3 months (2 trials, MD −7.23, 95% CI −11.72 to −2.74, I²=43%), and 12 months (2 trials, MD −3.31, 95% CI −6.60 to −0.02, I²=12%) versus the other treatment alone, combination therapy was also associated with better function at 1 month, (2 trials, SMD −0.40, 95% CI −0.73 to −0.07, I²=0%), 3 months (2 trials, SMD −0.22, −0.38 to −0.06, I²=33%), and 12 months (2 trials, SMD −0.21, 95% CI −0.34 to −0.09, I²=0%). One trial not included in the systematic review reported results consistent with these findings (SOE: low).
• For radicular low back pain, one good-quality trial found spinal manipulation plus home exercise and advice associated with greater improvement in leg and back pain at 12 weeks versus home exercise and advice alone (mean differences about 1 point on a 0 to 10 scale), but effects were smaller (0.3 to 0.7 points) and no longer statistically significant at 52 weeks (SOE: low).
• Harms were not reported well in most trials of spinal manipulation. No serious adverse events were reported and most adverse events were related to muscle soreness or transient increases in pain (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included 12 systematic reviews^{361, 425-436} with a total of 69 individual trials of spinal manipulation versus sham, an inactive treatment, or another active treatment for acute and chronic low back pain.¹⁶ The APS/ACP review concluded that there was fair evidence that spinal manipulation was associated with moderate benefits for acute and chronic low back pain.

One of the reviews of spinal manipulation in the APS/ACP review was subsequently updated as separate good-quality reviews for acute low back pain⁴³⁷ and chronic low back pain (Table 10; Appendix Tables E27, F27)⁴³⁸ The acute low back pain review included 19 randomized trials; eight of these trials were not included in the APS/ACP review.^439-446 Sample sizes ranged from 36 to 323 participants (total sample=2674). About half of the trials restricted inclusion to patients with acute low back pain,^{442-445, 447-451} four included patients with a mix of acute and subacute back pain^{439, 441, 446, 452} and six included patients with acute to chronic low back pain.^{440, 453-457}

A separate review⁴³⁸ included 26 trials (sample sizes 29 to 1,334, total sample=6070 participants) of spinal manipulation for chronic low back pain; 18 of these trials were not included in the prior Cochrane review.^{185, 192, 196, 402, 458-471} Only eight trials restricted inclusion to patients with symptoms longer than 3 months.^{192, 196, 402, 458, 460, 464, 465, 467, 470} The remainder permitted inclusion of patients with nonchronic symptoms, but the mean duration of back pain was generally months to years in duration.

Six studies of acute low back pain were rated as low risk of bias^{441-443, 445, 446, 453} and nine studies of chronic low back pain were rated as low risk of bias.^{185, 192, 461, 462, 469, 472-474} Methodological shortcomings in the high risk of bias studies included unblinded design, unclear allocation concealment methods, incomplete followup, selective reporting of outcomes and in one study⁴⁷⁵ significant baseline differences among groups.

In the trials included in the systematic reviews, spinal manipulation was compared against a wide variety of interventions, including various sham or inert therapies (placebo antiedema gel, detuned short-wave diathermy, bed rest, detuned ultrasound, corset and transcutaneous muscle stimulation, sham SMT), or another active intervention (acupuncture, back school, educational back booklet with or without additional counseling, exercise therapy, myofascial therapy, massage, pain clinic, pharmacological/analgesic therapy, short-wave diathermy, standard medical care [including analgesic therapy and advice/reassurance], standard physiotherapy, and ultrasound). The primary type of thrust technique used in the spinal manipulation interventions also varied. High-velocity low-amplitude (HVLA) thrust was used in most studies, though a combination of manipulation and mobilization or other mobilization techniques such as flexion-distraction or the Maitland method were used in 8 trials and unspecified types of SMT were used in 14 trials.

The number and frequency of manipulation treatments also varied among trials that reported this information. Approximately half of the acute low back pain trials did not report number of treatments, but those that did reported 1 and 10 treatment sessions. For chronic low back pain trials, the average maximum number of treatments allowed was 8 and the average duration of treatment 7 weeks in trials that provided this information. In both acute and chronic low back pain trials, followup ranged from 2 weeks to 2 years, with approximately half of the studies only reporting short-term outcomes (<3 months). One study of SMT for acute low back pain only measured the immediate effect of treatment, 2 days after the end of treatment.⁴⁴⁶

We identified 16 additional trials published subsequent to the updated reviews, 8 of which were trials for chronic low back pain,^{241, 476-482}two for acute or subacute low back pain^483-485, one for mixed duration low back pain⁴⁶⁶ one for radicular low back pain.⁴⁸⁶ (Table 18; Appendix Tables E28, F28). We also included 3 trials that were excluded from the systematic reviews because they enrolled patients with sciatica/radiculopathy.^487-489 Two studies were poor quality^{487, 488} and one was good quality.⁴⁸⁹ These additional trials varied in terms of the comparators including epidural steroid injections,⁴⁸⁷ chemonucleolysis,⁴⁸⁸ McKenzie,^{466, 481} physical therapy,⁴⁷⁸ active exercise^{241, 486} usual care,⁴⁸³ and inactive or sham treatments including detuned ultrasound,⁴⁷⁶ simulated manipulation,⁴⁸⁹ side lying,⁴⁷⁷ and light massage.⁴⁸⁰ The duration of treatment ranged from a single treatment⁴⁷⁹ on 1 day to 18 sessions over a period of 9 months.

Table 18

Characteristics and conclusions of included spinal manipulation trials.

Key Points

• For chronic low back pain, a systematic review found no difference between ultrasound versus sham ultrasound in pain at the end of treatment (3 trials, mean difference −7.12 on 0 to 100 scale, 95% CI −18.0 to 3.75, I²=77%) and two trials found no effects on pain 4 weeks after the end of treatment. Evidence from 5 trials was too inconsistent to determine effects on function, though a larger, good-quality trial found no effect on the RDQ (SOE: low for pain, insufficient for function).
• For chronic low back pain, a systematic review found no differences between ultrasound versus no ultrasound in pain (2 trials, mean difference −2.16, 95% CI −4.66 to 0.34, I²=0%) or back-specific function (2 trials, mean difference −0.41, 95% CI −3.14 to 2.32), but estimates were imprecise (SOE: low for pain and function).
• For chronic low back pain, evidence from 3 trials was insufficient to determine effects of ultrasound plus exercise versus exercise alone on pain or function, due to imprecision and methodological shortcomings (SOE: insufficient).
• For radicular low back pain due to spinal stenosis, a small trial found no differences between ultrasound plus exercise versus sham ultrasound plus exercise in back pain, leg pain, or the ODI after 3 weeks of therapy (SOE: insufficient)
• There was insufficient evidence from three small trials with methodological shortcomings to determine effects of ultrasound versus other interventions (SOE: insufficient).
• For radiculopathy, there was insufficient evidence from two small trials with methodological shortcomings to determine effects of ultrasound versus other interventions (SOE: insufficient).
• No study evaluated the effectiveness of ultrasound for acute nonradicular low back pain.
• One trial found no differences between ultrasound versus sham ultrasound in risk of any adverse event (6.0% vs. 5.9%, RR 1.03, 95% CI 0.49 to 2.13) or serious adverse events (1.3% vs. 2.7%, RR 0.48, 95% CI 0.12 to 1.88) (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included three small (n=15 to 73) trials of therapeutic ultrasound for low back pain.^494-496 All trials had methodological shortcomings. One trial (n=73) found ultrasound associated with a higher likelihood of being pain-free versus sham ultrasound or analgesics (41% vs. 12% vs. 6.8%, p<0.001 for ultrasound versus placebo), but used a quasi-randomized design (alternate allocation).⁴⁹⁵ In addition, all patients were placed on bed rest, a treatment no longer recommended. One small (n=10) trial of patients with chronic low back pain found ultrasound moderately superior to sham ultrasound after 10 treatment sessions, but had high loss to followup and did not perform intention-to-treat analysis.⁴⁹⁴ The third trial (n=36), which was published in 1960, found no difference between ultrasound and sham ultrasound for low back pain of unspecified duration in pain improvement after 1 month of therapy.⁴⁹⁶ The APS/ACP review concluded that there was insufficient evidence to determine effects of ultrasound.

We identified one good-quality systematic review published since the APS/ACP review (Table 10; Appendix Tables E29, F29).⁴⁹⁷ It included seven trials, including one of the trials described above⁴⁹⁴ and six subsequent trials.^{465, 494, 498-502} All trials enrolled patients with chronic nonradicular low back pain. Sample sizes ranged from 15 to 120 subjects. One trial was published in Croatian.⁵⁰² All studies evaluated 1 MHz continuous ultrasound at intensities from 1 to 2.5 W/cm², applied for 5 to 10 minutes or based on Gray's formula. The number of sessions ranged from 6 to 18. The review focused on outcomes immediately following the prescribed ultrasound treatment course; two trials also evaluated patients 4 weeks⁵⁰¹ and 6 months⁴⁶⁵ after the end of treatment. Four trials evaluated ultrasound versus sham ultrasound,^{494, 498, 501, 502} two trials ultrasound versus no ultrasound,^{499, 500} and three trials ultrasound versus other treatments (spinal manipulation,⁴⁶⁵ electrical stimulation,⁴⁹⁹ and phonophoresis⁵⁰⁰). In all of the trials except for one,⁴⁹⁴ patients in all treatment groups also underwent exercise therapy. Although all trials had methodological shortcomings, two trials^{498, 501} were classified by the systematic review as being at low risk of bias, based on meeting six or more of 12 risk of bias criteria. Patients were blinded to receipt of ultrasound in four trials,^{494, 498, 501, 502} care providers were blinded in none of the trials, and intention-to-treat analysis was reported in two trials.^{498, 501}

The largest randomized trial (n=455) of ultrasound was published after the systematic review (Table 19; Appendix Tables E30, F30).⁵⁰³ It compared ultrasound versus sham ultrasound for chronic nonradicular low back pain and evaluated patients 4 weeks after the end of treatment. About 20 percent of patients also underwent exercise therapy. The trial used a 2 × 2 factorial design in which patients were also randomized osteopathic manual treatment versus no manual treatment; there was no interaction between the ultrasound and manual treatment interventions. We also identified three small trials of ultrasound that were not included in the systematic review. One fair-quality trial (n=30) compared ultrasound versus low-level laser therapy for back pain ≥3 weeks,⁵⁰⁴ one fair-quality trial (n=45) compared ultrasound plus exercise, sham ultrasound plus exercise, and no treatment for spinal stenosis,⁵⁰⁵ and one poor-quality trial (n=60) of ultrasound versus traction or low-level laser therapy for acute radiculopathy due to herniated disc.⁵⁰⁶

Table 19

Characteristics and conclusions of included ultrasound trials.

Key Points

• For acute or subacute low back pain, evidence from single trials with methodological shortcomings was too limited to permit reliable conclusions regarding effectiveness (SOE: insufficient).
• For chronic low back pain, a systematic review found no differences between transcutaneous electrical nerve stimulation (TENS) versus sham TENS in pain intensity (4 trials, WMD −4.47 on a 0 to 100 scale, 95% CI −12.84 to 3.89) or function (2 trials, WMD −1.36 on a 0 to 100 scale, 95% CI −4.38 to 1.66) at short-term followup; most trials found no effect on pain or function at the end of a course of treatment (SOE: low for pain and function).
• For chronic low back pain, a systematic review found no differences between TENS versus acupuncture for short- (4 trials; SMD 0.15, 95% CI −0.33 to 0.63) or long-term pain (2 trials; SMD 0.32, 95% CI −0.33 to 0.96). Evidence for TENS versus other interventions was too limited to permit reliable conclusions (SOE: low for TENS vs. acupuncture).
• Evidence on harms associated with TENS was limited, but suggests an increased risk of skin site reactions without an increased risk of serious adverse events (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included a good-quality systematic review⁵⁰⁷ of TENS versus sham that included one good-quality trial,⁵⁰⁸ and one poor-quality trial.⁵⁰⁹ The first (good quality) trial (n=145) compared 4 weeks of treatment with followup 2 months after treatment cessation.⁵⁰⁸ The second trial (n=30) compared 2 weeks of TENS versus sham, with no post-treatment followup, and was rated poor quality due to lack of blinding, unclear allocation concealment, and incomplete outcome data.⁵⁰⁹ The APS/ACP review also included evidence from systematic reviews of acupuncture,³⁶³ massage,⁵¹⁰ spinal manipulation,⁴²⁶ traction, and superficial heat and cold⁵¹¹ that each included one to five trials of TENS versus these interventions. The APS/ACP review concluded that there was insufficient to determine the effects of TENS for acute or chronic low back pain.

We identified a more recent good-quality systematic review of TENS versus sham TENS⁵¹² that included the good-quality trial described above⁵⁰⁸ and four other trials (Table 10; Appendix Tables E31, F31).^{508, 513-516} The poor-quality study discussed above⁵⁰⁹ was not included, presumably because there was no post-treatment followup. The studies enrolled between 50 and 324 patients with chronic low back pain. Duration of treatment ranged from 2 to 8 weeks in five trials^{508, 513, 514, 516} and duration of followup ranged from 2 to 11 weeks in three trials;^{508, 513, 516} duration of followup was not reported in one crossover study.⁵¹⁵ One trial also included an exercise comparison group and one trial⁵¹³ included a PENS comparison group. Three trials were classified as higher-quality^{508, 514, 516} based on meeting at least six of eleven Cochrane Back Review Group criteria; common methodological shortcomings were failure to repeat adequate allocation concealment techniques, unblinded design or unclear blinding status, and failure to report intention-to-treat analysis.

We identified three other trials of TENS for chronic low back pain that compared TENS versus sham TENS^{517, 518} or interventional therapy⁵¹⁹ (Table 20; Appendix Tables E32, F32). Two fair-quality trials (n=236⁵¹⁷ and n=21⁵¹⁸), one of which included patients with or without radicular symptoms⁵¹⁷ evaluated TENS (3 months or 5 weeks of treatment) versus sham TENS. The third, good-quality trial (n=150), evaluated 2 weeks of TENS treatment versus interferential therapy.⁵¹⁹ All three trials only evaluated outcomes at the end of treatment.

Table 20

Characteristics and conclusions of included transcutaneous electrical nerve stimulation (TENS) trials.

Key Points

• There was insufficient evidence from five RCTs to determine effects of electrical muscle stimulation plus exercise versus exercise alone or versus other interventions, due to methodological limitations and imprecision (SOE: insufficient).
• There was insufficient evidence to determine harms of electrical muscle stimulation (SOE: insufficient).

Detailed Synthesis

The APS/ACP review did not evaluate effects of electrical muscle stimulation for low back pain. We identified five trials on the effects of electrical muscle stimulation for low back pain (Table 21; Appendix Tables E33, F33).^{492, 499, 524-526} The sample size ranged from 28 to 80 in four trials and was 164 in the fifth trial.⁴⁹² Four trials enrolled patients with chronic low back pain and the fifth trial⁴⁹² enrolled patients with back pain of 3 weeks to 6 months in duration. Two trials compared electrical muscle stimulation plus exercise versus exercise,^{499, 524} one trial electrical muscle stimulation plus exercise versus sham stimulation plus exercise,⁵²⁵ and three trials electrical muscle stimulation versus other interventions (ultrasound,⁴⁹⁹ TENS or sham TENS,⁵²⁶ and massage, manipulation, or lumbar supports⁴⁹²), with or without exercise. The duration of stimulation sessions ranged from 15 minutes to at least 8 hours, the number of sessions ranged from 2 to 60, and the duration of treatment ranged from 2 days to 2 months. The technical parameters of the stimulation varied. Outcomes were assessed at the end of 2 days to 8 weeks of therapy in four trials and at 6 months (4 months after the end of therapy) in the fifth trial.⁵²⁵ One trial was rated fair quality⁴⁹² and the other four poor quality. Methodological shortcomings included unclear randomization and allocation concealment methods, unblinded design, and lack of intention-to-treat analysis. In two trials, some subscales of the SF-36 were analyzed as mean differences and others as median differences without a rationale.^{499, 524}

Table 21

Characteristics and conclusions of included electrical muscle stimulation trials.

Key Points

• There was insufficient evidence from seven trials to determine effects of percutaneous electrical nerve stimulation (PENS) versus sham, PENS plus exercise versus exercise alone, or PENS versus other interventions, due to methodological limitations, inconsistency and imprecision (SOE: insufficient).
• Harms were poorly reported in trials of PENS (SOE: insufficient).

Detailed Synthesis

Percutaneous electrical nerve stimulation (PENS) involves the application of an electrical nerve stimulus via needles placed at the dermatomal levels corresponding to the pain (rather than at acupuncture sites). The APS/ACP review²⁹ included four trials (n=34 to 64) of percutaneous electrical nerve stimulation (PENS) for low back pain.^{513, 527, 528} Two trials compared PENS versus sham PENS,^{513, 527} one trial compared PENS plus physical therapy versus sham PENS plus physical therapy,⁵²⁸ three trials compared PENS versus TENS,^{513, 527, 529} and one trial compared PENS versus exercise.⁵¹³ Two trials enrolled patients with nonradicular low back pain,^{513, 528} one trial enrolled patients with radicular back pain,⁵²⁷ and one trial did not specify presence or absence of radicular symptoms.⁵²⁹ All trials enrolled patients with chronic low back pain; although the trial of radicular back pain enrolled patients with symptoms for >6 weeks, the mean duration was 21 months.⁵¹³ PENS was administered two or three times weekly for 2 to 8 weeks, with each session 15 to 45 minutes in duration. Outcomes were assessed at the end of treatment in two trials,^{513, 527} and 8 weeks⁵²⁹ or 3 months⁵²⁸ after the end of treatment. The APS/ACP review concluded that there was insufficient evidence to determine the effectiveness of PENS for acute or chronic low back pain.

We also identified three trials not included in the APS/ACP review (Table 22; Appendix Tables E34, F34). One trial (n=200) compared PENS (two 30-minute sessions weekly for 6 weeks), PENS plus supervised and home exercise, control (minimal) PENS plus exercise, and control PENS alone. One trial (n=112) compared PENS (three 30-minute sessions weekly for 3 weeks) versus dry needling for chronic low back pain.⁵³⁰ The other trial (n=75) compared different durations of PENS therapy (15, 30, or 45 minutes 3 times a week for 2 weeks) versus sham PENS (insertion of needles without stimulation).⁵³¹ Outcomes were assessed at the end of treatment in both trials.

Table 22

Characteristics and conclusions of included percutaneous electrical nerve stimulation (PENS) trials.

Among all trials, two trials were rated fair quality^{528, 532} and the rest were rated poor quality. Methodological shortcomings included inadequate description of randomization and allocation concealment methods, failure to report attrition, unblended or unclearly blinded design and failure to report intention-to-treat analysis. Three trials with a crossover design had a 1 week washout between treatments, but did not evaluate for potential carryover effects.^{513, 527, 531}

Key Points

• There was insufficient evidence from four trials to determine effects of interferential therapy versus other interventions, or interferential therapy plus another intervention versus the other interventions lone, due to methodological limitations and imprecision (SOE: insufficient).
• No study evaluated harms of interferential therapy (SOE: insufficient).

Detailed Synthesis

The APS/ACP review²⁹ included three trials (n=151 to 240) of interferential therapy for low back pain.^533-535 No trial compared interferential therapy versus sham therapy. One trial each compared interferential therapy versus spinal manipulation⁵³³ or traction.⁵³⁵ One of these trials also compared interferential therapy versus the combination of interferential therapy plus spinal manipulation.⁵³³ The third trial compared interferential therapy applied to the painful area versus to the area of the spinal nerve, each in combination with a self-care book, as well as against the self-care book alone.⁵³⁴ The trials focused on patients with nonradicular low back pain. The duration of symptoms was 4 to 12 weeks in two trials^{533, 534} and unspecified (mainly chronic) in the third.⁵³⁵ The trials varied in the number (range 3 to 10) and duration (10 to 30 minutes) of interferential therapy sessions and in technical parameters. Outcomes were assessed at 3 to 12 months (1 week to 10 months following the end of therapy). All of the trials were rated poor quality; methodological shortcomings included failure to blind patients or care providers, high attrition, and failure to perform intention-to-treat analysis; one trial⁵³⁴ also reported potentially important baseline differences. The APS/ACP review concluded that there was insufficient to determine the effectiveness of interferential therapy for acute or chronic low back pain.

One trial (n=62) published subsequent to the APS/ACP review evaluated interferential therapy versus superficial massage for chronic low back pain (Table 23; Appendix Tables E35, F35).⁵³⁶ Interferential therapy was administered for 30 minutes in 20 sessions over 10 weeks, with outcomes assessed at the end of therapy. The trial was rated fair quality; methodological shortcomings included unblinded design and failure to report use of cointerventions and compliance to assigned therapies.

Table 23

Characteristics and conclusions of included interferential therapy trials.

Key Points

• For acute or subacute low back pain, a systematic review found a heat wrap more effective than placebo for pain relief at 5 days (2 trials, mean difference 1.06 on a 0 to 5 scale, 95% CI 0.68 to 1.45) and disability at 4 days (mean difference −2.10 on the RDQ, 95% CI −3.19 to −1.01). Two subsequent trials also found a heat wrap associated with decreased pain intensity at 3 to 4 days (differences 16 to 20 points on a 0- to 100-point VAS) or increased pain relief at 8 hours (difference ∼1.5 points on a 0 to 5 scale). Another trial found a heat wrap during emergency transport associated with substantially lower pain intensity versus an unheated blanket upon arrival to the hospital (SOE: moderate for pain and function).
• For acute low back pain, one higher-quality trial found heat plus exercise associated with greater pain relief at day 7 (mean difference 1.40 on 0 to 10 scale, 95% CI 0.69 to 2.11) and on the RDQ (mean difference −3.20 on the RDQ, 95% CI −5.42 to −0) versus exercise without heat (SOE: low).
• One fair-quality trial found heat plus an NSAID associated with better pain scores versus an NSAID without heat at day 15, based on the McGill Pain Questionnaire (scoring methods unclear) (SOE: insufficient).
• For acute or subacute low back pain, a systematic review included one trial that found heat more effective for pain relief than acetaminophen (mean difference 0.90 on a 0 to 10 scale, 95% CI 0.50 to 1.30) or ibuprofen (0.65, 95% CI 0.25 to 1.05) after 1 to 2 days of treatment; the heat wrap was also associated with greater improvement on the RDQ (mean differences 2.00 on a 0 to 24 scale, 95% CI 0.86 to 3.14 and 2.20, 95% CI 1.11 to 3.29, respectively) (SOE: low for pain and function).
• For acute low back pain, a systematic review included one trial that found no clear differences between heat versus exercise in pain relief or function (SOE: low).
• No study compared superficial cold versus placebo or no cold treatment.
• For acute low back pain, one small trial with methodological shortcomings found cold plus naproxen associated with better pain scores versus naproxen alone, based on the McGill Pain Questionnaire (methods for scoring unclear). (SOE: insufficient).
• There was insufficient evidence from three trials to determine effects of heat versus cold, due to methodological limitations and imprecision (SOE: insufficient).
• Heat was not associated with increased risk of skin flushing versus no heat or placebo in two trials; no serious adverse events were reported with use of heat (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included a good-quality systematic (Cochrane) review⁵¹¹ of nine controlled clinical trials on the effects of heat or cold on low back pain. An updated version of the review was published in 2011, but included no additional trials (Table 2; Appendix Tables E36, F36).⁵³⁷ Of the studies included in the systematic review, five were randomized parallel-group trials.^{42, 538-541} The other four used alternate allocation or did not describe the allocation method;^{521, 542-544} one was a parallel group trial⁵⁴² and the other three used a crossover design.^{521, 543, 544} Four trials evaluated hot packs or heated wraps versus placebo or nonheated wraps,^{42, 539-541, 544} one trial heat plus exercise versus heat or exercise alone,⁵³⁸ one trial heat versus ibuprofen or acetaminophen,⁴² and two trials hot packs versus ice massage.^{542, 543} One trial⁵²¹ compared ice massage versus TENS and is discussed in the TENS section of this report. The sample sizes ranged from between 36 and 371 participants with acute pain (1 trial); mixed acute and subacute pain (4 trials); chronic (3 trials) or mixed acute pain; and subacute and chronic (1 trial) pain. Duration of treatment was generally 1 week or less, with followup no longer than 1 or 2 days after the end of treatment. In one trial, treatment duration was 25 to 27 minutes, with immediate post-treatment followup.⁵⁴¹

The systematic review rated the five RCTs^{42, 538-541} higher quality, based on meeting at least 6 of 11 Cochrane Back Review Group criteria,⁵⁴⁵ and the remaining studies were rated lower quality.^{521, 542-544} All studies had methodological limitations, including unblinded design and failure to adequately report methods of randomization and allocation concealment.⁵¹¹ Four trials reported funding by manufacturers of heat wraps.

We identified two fair-quality trials (n=30 and 51) published subsequent to the systematic review of heat therapy (4 hours daily for 4 days, 8 hours for 1 day) versus no heat therapy (Table 24; Appendix Tables E37, F37).^{546, 547} One trial⁵⁴⁷ enrolled patients with acute low back pain and the other⁵⁴⁶ included patients with acute or subacute low back pain. Methodological shortcomings included inadequate description of randomization and allocation concealment methods and nonblinded design; one trial⁵⁴⁶ also had high (21%) attrition. Both trials were funded by a manufacturer of heat wraps. We also identified a fair-quality trial (n=87) that compared superficial heat (hot water bottle 20 minutes twice daily for 1 week) plus naproxen 500 mg bid, superficial cold (ice 20 minutes twice daily for 1 week) plus naproxen, or naproxen alone for acute low back pain⁵⁴⁸ and a small (n=43), poor-quality trial of patients with acute low back pain in an occupational health setting evaluated3 days of heat-wrap therapy plus education versus education only through 14 days.⁵⁴⁹ Methodological shortcomings included inadequate description of randomization and allocation concealment, failure to report attrition, and unblinded design.

Table 24

Characteristics and conclusions of included superficial heat or cold trials.

Key Points

• For acute low back pain, there was insufficient evidence from one trial to determine effectiveness of low-level laser therapy versus sham laser, due to serious methodological shortcomings and imprecision (SOE: insufficient).
• For chronic low back pain, three of four trials found low-level laser therapy more effective than sham laser for pain, though methods for assessing pain and duration of followup varied; two trials found low-level laser therapy more effective than sham laser for function, with small magnitude of effects (SOE: low for pain and function).
• For acute or subacute low back pain, one trial found low-level laser therapy plus an NSAID associated with lower pain intensity versus sham laser plus an NSAID or the NSAID alone (mean differences 9 to 14 points on a 0 to 100 VAS); effects on the ODI also favored combination treatment but were smaller (differences <6 points) (SOE: low).
• For chronic low back pain, there was insufficient evidence from three trials to determine effects of low-level laser therapy plus exercise versus the other sham laser plus exercise alone, due to methodological shortcomings and inconsistency (SOE: insufficient).
• There was insufficient evidence to determine effects of low-level laser therapy versus another intervention, due to methodological shortcomings and imprecision (SOE: insufficient).
• There was insufficient evidence to determine effects of different wavelengths of low-level laser therapy or different doses, due to methodological limitations and imprecision (SOE: insufficient).
• Harms were not well-reported in trials of low-level laser therapy, but no serious adverse events and no harms were reported (SOE: low)

Detailed Synthesis

Low level laser therapy involves administration of a single wavelength of light (usually from 632 to 904 nm) that does not emit heat but may affect underlying connective tissues and have potential anti-inflammatory effects. The APS/ACP review²⁹ included seven trials^550-556 of low-level laser therapy for low back pain. Four trials were conducted in patients with chronic low back pain, one trial in patients with acute low back pain, and two trials did not specify the duration of symptoms. The APS/ACP review found insufficient evidence to determine effectiveness of low-level laser therapy versus sham or other interventions, due to variability across trials in terms of laser types and doses, outcomes, duration of followup, and inconsistency in results. A recent systematic review⁵⁵⁷ included six^550-555 of the trials included in the APS/ACP review plus one additional trial (Table 2).⁵⁵⁸ We also identified five recently published trials not included in prior reviews (Table 25; Appendix Tables E38, F38).^559-563

Table 25

Characteristics and conclusions of included low-level laser therapy trials.

In total, after excluding one poor-quality trial with uninterpretable findings,⁵⁵⁶ 10 trials assessed low-level laser therapy for low back pain.^{550-555, 558-563} Laser wavelengths ranged from 830 to 10600 nm and five of the trials used a 904 nm laser.^{551, 552, 555, 560, 562} Duration of treatment ranged from 1 day to 6 weeks, followup was from 1 day to 1 year, and there were 6 to 20 laser treatment sessions (1 trial⁵⁵⁵ did not report the treatment protocol and 1 trial⁵⁵³ assessed outcomes after a single treatment). Five trials compared laser versus sham^{550-553, 559} and two trials compared different laser doses, either with⁵⁵⁵ or without⁵⁶² a sham group. The other five trials compared laser plus another treatment (heat, exercise, or NSAID) versus the other treatment, either alone or in combination with sham laser.^{554, 558, 560, 561, 563} Sample sizes in nine of the trials ranged from 20 to 120 and in the tenth (largest) study was 546.⁵⁶⁰ Patients had chronic low back pain in seven trials^{550-552, 554, 558, 559, 563} and acute low back pain in three trials.^{555, 560, 562} One trial enrolled both subacute and chronic pain patients⁵⁶¹ and one trial did not report the duration of pain.⁵⁵³ Two trials^{560, 561} were rated good quality, two^{554, 555} poor quality, and eight trials fair quality. Methodological limitations in the fair- and poor-quality studies included inadequate reporting of treatment allocation, unblinded design, use of cointerventions, unclear or low compliance, and high or unclear attrition.

Key Points

• For back pain of mixed duration, there was insufficient evidence from five RCTs to determine effects of short-wave diathermy versus sham diathermy, due to methodological limitations and imprecision (SOE: insufficient).
• No study evaluated harms of short-wave diathermy.

Detailed Synthesis

The APS/ACP review²⁹ included three trials (n=24 to 400) of short-wave diathermy for low back pain.^{450, 475, 565} Two trials^{475, 565} compared short-wave diathermy versus sham diathermy and all three trials compared short-wave diathermy versus other interventions (exercises,⁵⁶⁵ traction,⁵⁶⁵ or manipulation^{450 , 475}). The trials focused on patients with nonradicular low back pain. One trial enrolled patients with acute (<3 weeks) back pain,⁴⁵⁰ one trial patients with low back pain for 2 to 12 months,(Gibson) and the third patients with back pain for >1 week.⁵⁶⁵ The trials varied in the number of sessions (6, 12, or not specified); only one trial⁵⁶⁵ specified the duration of each session (20 minutes). Outcomes were assessed at the end of a 2-week course of therapy in two trials^{450, 565} and at 12 weeks (8 weeks after the end of therapy) in the third trial.⁴⁷⁵ Two trials^{475, 565} were rated fair quality and one trial⁴⁵⁰ poor quality; methodological shortcomings included unclear randomization and allocation concealment methods, failure to blind care providers and outcome assessors, and failure to report use of cointerventions and compliance. The poor-quality trial also did not blind patients. The APS/ACP review found insufficient evidence to determine effects of short-wave diathermy for acute or chronic low back pain.

Two trials (n=97 and 102) published subsequent to the APS/ACP review evaluated short-wave diathermy versus sham diathermy for chronic low back pain (Table 26; Appendix Tables E39, F39).^{566, 567} Short-wave diathermy was administered for 15 minutes in 18 sessions over 6 weeks, with outcomes assessed at the end of therapy. Both trials were rated poor quality; methodological shortcomings included unclear randomization and allocation concealment methods, failure to report attrition, lack of intention-to-treat analysis, and failure to blind caregivers and outcome assessors.

Table 26

Characteristics and conclusions of included diathermy trials.

Key Points

• For acute or subacute low back pain, there was insufficient evidence from five trials to determine effects of lumbar supports versus no lumbar supports or an inactive treatment, due to methodological shortcomings and inconsistent results (SOE: insufficient).
• For chronic low back pain, there was insufficient evidence from two trials to determine effects of lumbar supports versus no lumbar supports, due to methodological shortcomings and inconsistent results (SOE: insufficient).
• For back pain of mixed duration, one trial found an inextensible but not extensible lumbar support associated with greater improvement in function versus no lumbar support, but effects were small. There was no clear effect on function (SOE: low).
• For acute or subacute low back pain, one trial found no differences between a lumbar support plus an education program versus an education program alone in pain or function after 1 year (SOE: low for pain and function).
• For chronic low back pain, one trial found no difference between a lumbar support plus exercise (muscle strengthening) versus exercise alone in short-term (8 weeks) or long-term (6 months) pain or function (SOE: low for pain and function).
• Three trials found no clear differences between lumbar supports versus other active treatments in pain or function (SOE: low for pain and function).
• There was insufficient evidence from 3 trials to determine comparative effects of different types of lumbar supports for chronic low back pain or back pain of mixed duration, due to heterogeneous comparisons, methodological shortcomings and imprecision (SOE: insufficient).
• Trials reported no harms associated with use of lumbar supports (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included a good-quality systematic review⁴⁴ with six trials^{39, 568-572} of lumbar supports for low back pain. Sample sizes ranged from 19 to 456 subjects. One of the trials was classified as high quality.⁵⁶⁹ The APS/ACP review concluded that there was insufficient evidence to determine effects of lumbar supports for acute or chronic low back pain.

An updated version of the systematic review⁵⁷³ with two additional trials^{574, 575} (eight total) has since been published (Table 2; Appendix Tables E40, F40). Six trials compared lumbar supports versus no lumbar supports,^{39, 568, 569, 571, 572, 575} three trials lumbar supports versus other active interventions (e.g., spinal manipulation therapy, exercise, massage),^{39, 568, 569, 575} and two trials compared different types of lumbar supports^{570, 574} The types of lumbar supports included flexible and semi-rigid corset made of various materials and a pneumatic lumbar support. Duration of treatment ranged from 3 weeks to 2 months. Trials evaluated patients at the end of treatment; three trials^{39, 568, 574} also evaluated patients 6 to 16 months after the end of treatment. Three trials enrolled patients with chronic pain^{570, 574, 575} and four enrolled patients with low back pain of mixed duration;^{39, 568, 569, 572} one trial⁵⁷¹ did not report the duration of pain. Sample sizes ranged from 19 to 456 (total n=1,361). All of the studies except for one⁵⁶⁹ were rated lower quality by the systematic review, based on meeting at least 5 of 10 Cochrane Back Review Group criteria.⁵⁴⁵ Methodological shortcomings in the trials included failure to describe adequate randomization and allocation concealment methods, lack of blinding of outcome assessors, inadequate or unclear compliance, and possible differential use of cointerventions.⁵⁷³

We identified three trials (n=50, 98 and 217) published since the updated Cochrane review of lumbar supports versus no support for subacute,⁵⁷⁶ chronic, ⁵⁷⁷ or mixed duration low back pain⁵⁷⁸ and one trial (n=433) of lumbar supports plus education versus education alone for acute or subacute low back pain⁵⁷⁹ (Table 27; Appendix Tables E41, F41). One of the trials also compared two types of lumbar supports (inextensible [stiffer] versus extensible).⁵⁷⁸ Treatment duration was 2 weeks, 3 months, or 6 months in three trials, with followup up through the end of treatment. The fourth trial did not report treatment duration, but followed patients for 1 year. All four trials were rated fair quality; methodological limitations included unclear allocation concealment methods, unclear compliance and unblended design.

Table 27

Characteristics and conclusions of included lumbar support trials.

Key Findings

• For low back pain with or without radicular symptoms, a systematic review included 13 trials that found no clear differences with inconsistent effects of traction versus placebo, sham, or no treatment in pain, function, or other outcomes, though two trials reported favorable effects on pain in patients with radicular back pain (SOE: insufficient for pain and function).
• For low back pain with or without radicular symptoms, a systematic review included five trials that found no clear differences between traction versus physiotherapy versus physiotherapy alone (SOE: low).
• For low back pain with or without radicular symptoms, a systematic review included 15 trials of traction versus other interventions that found no clear between traction versus other active interventions in pain or function (SOE: low for pain and function).
• A systematic review included five trials that found no clear differences between different types of traction (SOE: low).
• Eleven trials of traction in a systematic review reported no adverse events or no difference in risk of adverse events versus placebo or other interventions. Three subsequent trials reported findings consistent with the systematic review (SOE: low).

Detailed Synthesis

The APS/ACP review²⁹ included a large, good-quality systematic review⁵⁸⁰ with 23 trials of traction versus sham or no treatment, sham versus other interventions, or one type of traction versus another. The review was subsequently updated to include 32 trials with 2,762 patients (Table 2; Appendix Tables E42, F42).⁵⁸¹ Thirteen trials compared traction versus placebo, sham traction, or no treatment; 15 trials compared traction versus other active treatments (including exercise [6 trials], heat therapy [2 trials], or other passive physical modalities [7 trials]); and five trials compared one type of traction versus another. Ten trials assessed participants with chronic pain and participants with subacute pain were included in one trial. In the remaining trials, duration of pain was mixed (17 trials) or not clearly reported (5 trials). Of the 32 included trials, 23 enrolled populations with radicular low back pain. Among the other nine studies, eight enrolled a mixed population with radicular and nonradicular pain and one enrolled only patients with nonradicular low back pain. Duration of followup ranged from 1 week to 1 year. Only three studies^{460, 582, 583} reported outcomes beyond 6 months followup. Sixteen of the 32 included trials were judged to have a low risk of bias (i.e., quality score ≥6/12.)

We also identified three trials (in four publications) not included in the updated systematic review (Table 28; Appendix Tables E43, F43).^584-587 Each trial (n=24 to 80) compared combination treatment with traction plus another active intervention versus the active intervention alone. None of the trials clearly stated the duration of low back pain in study participants, but inclusion criteria for two trials^584-586 required 3 months or more of pain at baseline (subacute/chronic) and the third⁵⁸⁷ required no more than 6 months of pain at baseline (acute/subacute). Two trials compared 10 weeks of traction in combination with usual care (infrared lamp and stretching^{584, 585}or hot packs and interferential therapy⁵⁸⁶) versus usual care alone. The third trial compared inversion traction plus physiotherapy with physiotherapy alone.⁵⁸⁷ Study participants received treatment for 10 weeks in two trials^584-586 and for 4 weeks in the third,⁵⁸⁷ with respective followup of 6 months and 56 weeks. Two trials were rated fair quality^584-586 and one poor quality.⁵⁸⁷ Methodological shortcomings included unblinded design and in the case of the poor-quality study,⁵⁸⁷ inadequate description of randomization and allocation concealment techniques, and incomplete followup.

Table 28

Characteristics and conclusions of included traction trials.

Key Points

• For chronic low back pain, three trials found no differences between a Kinesio Taping^® versus sham taping in back-specific function after 5 to 12 weeks; effects on pain were inconsistent (SOE: low for function, insufficient for pain).
• For chronic low back pain, there was insufficient evidence from 1 trial to determine effects of Functional Fascial Taping plus exercise versus sham taping plus exercise, due to methodological limitations and imprecision (SOE: insufficient).
• For chronic low back pain, two trials found no differences between Kinesio Taping^® versus exercise therapy in pain or function (SOE: low for pain and function).
• No trial of taping reported harms.

Detailed Synthesis

The APS/ACP review did not evaluate taping for low back pain. We identified six trials on the effects of taping (Table 29; Appendix Tables E44, F44).^592-597 Sample sizes ranged from 20 to 60 in five trials and was 148 in the fifth trial.⁵⁹⁷ Five trials evaluated Kinesio Taping^{®592, 593, 595-597} and one trial Functional Fascial Taping.⁵⁹⁴ Three trials of Kinesio Taping^{592, 593, 597} and the trial of Functional Fascial Taping^®594 evaluated a sham taping (taping applied without tension) comparison. In the Functional Fascial Taping trial, patients in both groups also received instruction in home trunk flexion exercises. Among the Kinesio Taping trials, one trial⁵⁹⁵ compared Kinesio Taping versus exercise therapy without Kinesio Taping, one trial ⁵⁹⁶ compared Kinesio Taping versus exercise therapy or the combination of taping and exercise, and one trial⁵⁹² compared Kinesio Taping plus physical therapy (hot pack, ultrasound, and TENS) versus sham taping plus physical therapy. The five trials of Kinesio Taping enrolled patients with chronic low back pain and the trial of Functional Fascial Taping enrolled patients with back pain for >6 weeks, though the median duration was 32 to 39 weeks. The taping techniques all involved some degree of tension, though the taping pattern, reapplication interval, and duration of treatment (7 days to 12 weeks) varied. Outcomes were assessed at the end of a 4-week course of therapy in two trials^{595, 596} and at 5 to 12 weeks (4 to 10 weeks after the end of therapy in three trials and at the end of therapy in one trial⁵⁹²). Two trials were rated good quality,^{593, 597} three fair quality,^{592, 594, 596} and one poor quality.⁵⁹⁵ Methodological shortcomings in the fair- and poor-quality trials included unclear randomization and allocation concealment methods, unblinded design, failure to report attrition, and unclear use of intention-to-treat analysis.

Table 29

Characteristics and conclusions of included taping trials.