Evidence-based Evaluation of Complementary Health Approaches for Pain Management in the United States

Abstract

While most pain is acute, resolving within a few days or weeks, millions of Americans suffer from persistent or recurring pain that may become chronic and debilitating. Medications may provide only partial relief from this chronic pain and can be associated with unwanted side effects. As a result, many individuals turn to complementary health approaches as part of their pain management strategy. This article examines the clinical trial evidence for the efficacy and safety of several specific approaches – acupuncture, manipulation, massage therapy, relaxation techniques including meditation, selected natural product supplements (chondroitin, glucosamine, methylsulfonylmethane, S-adenoysl methionine), tai chi, and yoga – as used to manage chronic pain and related disability associated with back pain, fibromyalgia, osteoarthritis, neck pain, and severe headaches or migraines.

INTRODUCTION

The most recent national estimate suggests that 126 million adults suffer from some pain in a given year¹ with about one-third (40 million adults) suffering from severe pain. Pain is often associated with poor general health, health-related disability and increased health care utilization.² Yet according to the Institute of Medicine (2011),² pain is routinely undertreated in health care; pain care that is provided is often fragmented, without a comprehensive assessment or treatment plan, and patients may encounter difficulty obtaining the full range of potential treatments.

Back pain, joint pain, neck pain and headaches are among the most common types of pain experienced by US adults (Table 1).^3–6 The prevalence rates for these conditions have remained relatively stable over time. Among the many pharmacological and non-pharmacological approaches that have been incorporated into pain management strategies are complementary health approaches. This broad category of care includes procedures by licensed providers such as acupuncturists, chiropractors, and massage therapists, as well as self-care approaches such as relaxation techniques (e.g., meditation), and meditative movement-based approaches (e.g., yoga and tai chi), as well as natural products such a glucosamine and herbal medicines. National surveys going back more than 25 years have consistently found that these complementary approaches are used by about 30% to 40% of the US public in a given year,^7–11 although use of a given approach may wax and wane over time (Table 2).

Table 1.

Age-adjusted percentages of selected health conditions among U.S. adults, NHIS^a 2002, 2007, 2012

	2002b	2007c	2012de
Low Back Pain / 3 months	26.4	25.4	27.6
Arthritis	20.9	20.3	20.6
Neck Pain / 3 months	13.8	12.8	13.9
Severe Headache or Migraine / 3 months	15	12.3	14.1
Fibromyalgia / lifetime			1.75

^aNational Health Interview Survey

^bLethbridge-Cejku et al. (2004)³

^cPleis et al. (2009)⁴

^dBlackwell et al. (2014) ⁵

^eWalitt et al. (2015)⁶

Table 2.

Age-adjusted percentages of use for selected complementary health approaches by U.S. adults, 2002–2012

	2002a	2007b	2012c
Acupuncture	1.1	1.4	1.5
Manipulation	7.5	8.6	8.4
Massage therapy	5.0	8.3	6.9
Meditation	7.6	9.4	8
Natural product supplements	18.9	17.7	17.7
Yoga, tai chi, and qi gong	5.8	6.7	10.1

^aBarnes et al. (2008)⁷,

^bBarnes et al. (2004)⁸

^cClarke et al. (2015)⁹

While a substantial part of this use is for overall wellness and prevention,^12,13 painful conditions are the most common health conditions for which individuals turn to these complementary approaches.^{7, 8, 10, 11, 14} In 2007, for example, about 14.3 million adults used a complementary health approach for their back pain, while about 5.0 million used these approaches for their neck pain, and 3.1 million for their arthritis.⁷ Far fewer individuals used complementary health approaches for other chronic diseases such as depression (1.0 million), hypertension (0.8 million), diabetes (0.7 million) or cancer (0.4 million).

Based on national survey data,¹⁴ this high use of complementary health approaches for painful conditions translated into $8.5 billion out-of-pocket for these approaches to manage back pain, $3.6 billion to manage neck pain, and $2.3 billion to manage arthritis. Substantially less is spent out-of-pocket on complementary health approaches to treat other chronic health conditions such as depression ($1.1 billion), hypertension ($0.7 billion), diabetes ($0.3 billion), and cancer ($0.2 billion).

Given the high use of complementary health approaches for pain, a number of specific complementary approaches have undergone mechanistic and clinical evaluations culminating in phase III trials. This article examines the clinical trial evidence for the efficacy, effectiveness and safety of several widely used approaches – acupuncture, manipulation, massage therapy, relaxation techniques including meditation, selected natural product supplements (e.g., chondroitin and glucosamine) tai chi, and yoga (defined in Supplemental Appendix 1) – as used to manage chronic pain and related disability associated with back pain, osteoarthritis, neck pain, and severe headaches or migraines, conditions frequently seen and managed in the primary care setting. Fibromyalgia was included in this review as an example of a complex pain syndrome that is often managed with a multi-modal approach that may include complementary approaches. Cancer pain is certainly a major public health concern but is more likely to be addressed outside of the primary care setting (e.g., by oncologists, at cancer centers, as part of palliative care, etc.).

The randomized, controlled clinical trial (RCT) is considered the strongest study design for investigating the efficacy, effectiveness and safety of pharmacological, behavior, and physical interventions. To identify examples of RCT’s for each complementary approach, we performed searches of the MEDLINE database for articles published from 1966 through March, 2016 using the search strategy outlined in Supplemental Appendix 2. In order to make this review as relevant as possible to the primary care provider in the US, we limit this review to RCT’s either done in the US or that included participants from the US. This decision is based on two factors. First, the unique health care system in the US versus other countries means that the standard care or usual care control groups used in the US and other countries may vary substantially. Thus whether or not a given complementary approach performs better than usual care in another country may not reflect how the approach would perform in US trials. Another factor is that the training and licensure of acupuncturists, chiropractors and naturopathic doctors vary substantially among countries, as does the marketing, regulation and use of dietary supplements. For instance, in Germany, the location of some of the largest acupuncture trials, acupuncture is only practiced by medical doctors, while the vast majority of acupuncture treatment in the US is provided by licensed acupuncturists. Thus, the findings from German trials may not be directly comparable to acupuncture as practiced in the US.

Brief summaries of the reviewed RCTs are presented below with details provided in online supplemental tables. The findings of these RCTs also illustrate several methodological issues that should be considered when interpreting the trial data. These issues are summarized briefly at the end of this article,

LOW BACK PAIN

Acupuncture

We found four RCT’s (total participants = 1,092)^15–18 that assessed the clinical benefit of acupuncture for treatment of low back pain (LBP) (age range: 28–60, mostly white race), and used primary study outcomes of self-report of pain intensity (numeric rating scale [NRS] or visual analog scale [VAS]) and/or functional disability (Roland-Morris Disability Questionnaire [RMDQ], Oswestry Disability index [ODI], or Disability Rating Index). Cherkin et al. (2001¹⁵ & 2009¹⁶) both reported modest improvement in pain intensity and function compared to usual care. In pregnant women using auricular acupuncture, Wang et al. (2009)¹⁷ found a significant reduction in pain intensity and function compared to no treatment control. Comparison of verum- to sham-acupuncture had mixed results with two RCTS^{16, 18} finding no significant difference and 1 RCT¹⁷ finding a slight but significant difference. No significant adverse events were reported.

Massage Therapy

We summarize seven RCTs studying the use of massage for LBP^{15, 19–25} (total participants = 829). Massage types included Swedish/relaxation, structural, structural integration, and muscle energy; sessions varied in duration from 15 – 90 minutes. For chronic LBP (cLBP), 2 larger studies^{15, 19} reporting massage compared to usual care resulted in modest improvements in pain and function at 10 weeks, but the benefit was not sustained at 52 weeks.¹⁵ Three smaller studies compared massage to either usual outpatient rehabilitation²⁴ or relaxation^{22, 23} and did not observe significant between-group differences for pain and/or function. For acute or subacute LBP, 2 smaller studies found significant, albeit modest, improvements in pain compared to no treatment (−1.5 pts. NRS) ²¹ or function (−18% ODI)²⁵ compared to a putative placebo. No RCTs reported significant adverse events.

Osteopathic Manipulative Therapy

Six RCTs of osteopathic manipulative therapy (OMT) for LBP are reviewed (total participants = 1308).^26–31 2 RCTs examined OMT compared to sham-OMT for cLBP^{30, 31} using similar intervention paradigms and reported mixed results with 1 not finding any significant difference and 1 showing a −9mm reduction in pain intensity (VAS). 2 RCTs compared OMT to usual care for acute or subacute LBP^{26, 27} and both reported no significant improvement for function/disability but mixed results for pain intensity for between-group differences. In pregnant women with LBP, 2 studies compared the benefit of adding OMT or a placebo-ultrasound to usual obstetric care, and reported significant, modest improvements with added care but no significant between-group differences.^{28, 29} No RCTs reported significant adverse events.

Spinal Manipulation

We reviewed 24 RCTs^32–55 (total participants = 4503, all adults) of spinal manipulation (SM) for LBP. Recent data suggest that for cLBP, the “dose” of SM (defined as the # of sessions) may affect outcomes,^{45, 46, 56} and hence the dose utilized was dichotomized for this report depending upon whether 6+ sessions of SM were provided in a given study. No RCTs reported significant adverse events.

Chronic LBP

There were 9 studies of SM for cLBP (total participants = 1882),^{32, 35–37, 40, 44–46, 50} 8 of which employed 6+ sessions for treatment, and 1 of these studies was principally (> 2/3) of adult women.³² For the 4 larger RCTs (each with 200+ participants)^{37, 44, 46, 50} that compared HVLA-SM to an active comparison (exercise, usual care, physical therapy, light touch), 3 found significant, albeit modest, between-group differences for pain intensity and/or function. 2 studies directly examined dose of HVLA-SM^{45, 46} and both found the higher doses (12 and 18 sessions of SM) provided larger improvement at 12 and 52 weeks, respectively. 1 RCT utilized a putative placebo-SM³² and found that after the first session that verum-SM > sham-SM > no treatment, but at 2 weeks, there were no significant between-group differences. 1 RCT compared thrust to non-thrust SM⁴⁰ (less than 6 sessions) and reported no significant between-group differences.

Acute, sub-acute, or mixed LBP

There were 15 studies of SM for either acute, subacute, or a mixture of types of LBP (total participants = 2621).^{33, 34, 38, 39, 41–43, 47–49, 51–55} Six of these studies were of modest – moderate size (i.e., > 100 participants) and utilized 6+ sessions of SM.^{38, 42, 49, 51, 53, 54} The results of these studies were mixed with some demonstrating modest, significant benefit of SM compared to active intervention (physical therapy, back school, medication, usual care) at ~4 weeks for pain intensity and/or function,^{42, 49, 54} but others not reporting significant between-group differences.^{38, 51, 53} 1 RCT examined pregnant women with LBP⁴² and found that adding SM + exercise to usual obstetric care provided modest improvement in pain and function/disability.

There were 5 studies of SM for either acute and/or subacute LBP that used less than 6 treatment sessions,^{33, 39, 41, 47, 55} but only 1 of these (and the largest) was an effectiveness study: Fritz et al. (2015)⁴¹ did a RCT of 4 sessions over 4 weeks of SM + exercise vs. usual primary care and found at 3 months significant between-group improvement in function (−3.2 pts ODI). The other 4 studies (sample sizes ranged from 54 – 123) compared different types of SM and/or SM to an active intervention (e.g., exercise). In general, these 4 studies reported that all groups improved and there were very small or no significant between-group differences; though it is unclear whether these 4 studies were sufficiently powered to definitively ascertain whether differences existed.

There were 4 studies of SM that by design had mixed LBP^{34, 48, 52, 53} (i.e., the inclusion criteria allowed participants with acute, sub-acute, or chronic LBP; 3 of these studies utilized 6+ treatment sessions. All studies had 1 or more active comparison groups (e.g., massage, electrical stimulation, usual care, steroid injection, physical therapy), and in all studies, all groups improved but there were no significant between-group differences.

Yoga

Six RCTs of yoga for cLBP^57–62 are examined (total participants = 596, all adults, predominantly female). Three named forms of yoga were studied: hatha,^{57, 58} viniyoga,^{59, 60} and Iyengar;^{61, 62} all were done in group settings, with class durations from 60 – 90 minutes, and number of sessions ranging from 12 – 24, either once or twice per week, with recommendations for home practice. Compared to usual care, 2 studies^{59, 60} found yoga provided improvements in pain bothersomeness and function, but the results were mixed when compared to exercise/stretching. A dose-response study⁵⁷ compared 1/week to 2/week classes and found both had equivalent improvements in pain intensity and function. 3 smaller studies compared yoga to wait-list^{58, 61} or education control⁶² and reported significant, modest reductions in pain intensity and function/disability. No RCTs reported significant adverse events.

Additional information on all back pain RCT’s can be found in Supplemental Table 1.

FIBROMYLGIA

All trials we reviewed for fibromyalgia used the 1990 American College of Rheumatology (ACR) classification criteria⁶³ except for one⁶⁴ that used an older definition.

Acupuncture

Four RCT’s examined acupuncture versus sham acupuncture for pain, physical function, global well-being, sleep, fatigue, and adverse events.^65–68 Martin et al. (2006)⁶⁸ found a significant improvement between the electro-acupuncture versus a sham electro-acupuncture control group. Differences were seen on the Fibromyalgia, Impact Questionnaire (FIQ), on fatigue and on anxiety. No other trial found significant differences between groups on any outcome. There were no serious adverse events reported in any of these studies. In one study, minor adverse events (e.g. discomfort at site of needle insertion or simulation of needles) were reported by 89% of participants.⁶⁵

Relaxation Techniques

Two studies (93 total participants, mostly female and white) looked at biofeedback versus control groups (attention control and placebo) as a treatment for symptoms of fibromyalgia.^{64, 69} Buckelew et al. (1998)⁶⁴ saw a significant improvement in the Tender Point Index in the biofeedback group versus an attention control but not for any other outcome measure. Nelson et al. (2010)⁶⁹ did not find any differences between biofeedback and a placebo biofeedback. A small study (90 women) found that MBSR significantly reduced perceived stress, sleep disturbance, and lessened the severity of symptoms in persons with fibromyalgia versus a waitlist control group.⁷⁰ Another RCT examined the effects of Affective Self-Awareness, a technique that places primary importance on the awareness and expression of emotions underlying fibromyalgia symptoms, on 45 women with fibromyalgia, and found significant pain reduction and improved physical functioning versus a waitlist control group.⁷¹ Astin et al. (2003)⁷² examined the effects of an intervention combining mindfulness meditation and qi gong and found that the combined intervention was no better than an educational/support control group for pain, depression, and physical functioning. Two studies^{73, 74} (112 total participants, mostly female) examined guided imagery versus usual care control as a treatment for symptoms of fibromyalgia. One study⁷⁴ found a significant decrease in the FIQ compared to the usual care control group. The second study⁷³ found positive effects of guided imagery on pain intensity, fatigue, and depression versus the control group. Both studies found improvements in self-efficacy for managing symptoms. Only one study⁶⁹ reported on adverse events and there were none noted.

Massage

A small study (12 women) examined Swedish massage versus myofascial release therapy for fibromyalgia symptoms.⁷⁵ No difference was seen between groups on the FIQ.

Tai Chi

A study of 98 adults with fibromyalgia, aged 40 and over, mostly white and female, compared Yang style tai chi (modified for fibromyalgia patients) with an educational control and found that the tai chi group had a greater decrease in the FIQ⁷⁶. Another study (59 adults with fibromyalgia) compared Yang style tai chi to a control combining wellness education and stretching classes, and found that the tai chi group had greater improvement in the FIQ.⁷⁷

Yoga

A small study (53 women) looked at yoga versus waitlist for management of fibromyalgia symptoms and found that those practicing yoga had significant improvement in the FIQ.⁷⁸ No adverse events were noted.

Additional information on all fibromyalgia RCT’s can be found in Supplemental Table 2.

NECK PAIN

Massage

Four randomized controlled trials examined whether massage could relieve symptoms associated with chronic neck pain.^79–82 One study did not report patient demographics, the others studied subjects ages 20 to 64. Primary outcomes included The Neck Disability Index (NDI), an 11 point neck bothersome score, the VAS pain scale, and range of motion (ROM). Sherman et al. (2009)⁸¹ found significant improvement on the NDI for those randomized to 10 massage therapy session over 10 weeks versus those assigned to a self-care book on managing neck problems. In an RCT by Field et al.,⁸⁰ individuals were randomized to either a wait list control, or 30 minutes of massage therapy weekly for 4 weeks combined with daily self-massage. At the completion of the intervention, those assigned to massage therapy showed improvements in pain and range of motion compared to the control group. Sherman et al.⁸² demonstrated a dose response relationship between the number and duration of message sessions per week and improvement in the NDI and neck pain intensity. The findings indicated that 60 minutes of massage for 2 to 3 times per week was significantly better than either 30 or 60 minutes of massage once per week after the 4 weeks of treatment. In a follow-up to Sherman et al. (2014),⁸² Cook et al. (2015)⁷⁹ re-consented and randomized participants to one additional massage therapy session per week for 6 additional weeks. Those randomized to the booster sessions had significantly improved pain and function verses those that did not receive the booster at the end of treatment; the difference between groups was no longer significant after 14 weeks of follow-up.

Spinal manipulation

We reviewed three randomized trials of spinal manipulation for neck pain.^83–85 One study assessed manipulation compared to mobilization with a 2×2×2 factorial design: with or without heat, or with or without electrical muscle stimulation (EMS);⁸⁴ no significant differences in outcomes were seen between groups. Evans et al.⁸³ compared spinal manipulation combined with supervised exercising to supervised exercising alone and also to home exercise. After completion of the 12-week intervention, no difference was seen between spinal manipulation combined with supervised exercising and supervised exercising alone; however, both these groups showed significant improvement in neck pain versus only home exercise. Maiers et al.⁸⁵ assessed the efficacy of three groups: 1) manipulation plus home exercise; 2) versus supervised rehabilitation exercise plus home exercise; and 3) home exercise alone. Spinal manipulation with home exercise produced significantly better reduction in pain than home exercise alone. No significant difference was seen between manipulation and home exercise, versus supervised rehabilitation exercise plus home exercise.

There was one RCT of manual cervical distraction,⁸⁶ a traction-based therapy with low, medium, and high forces assessed. The goal of the study was to identify a viable sham-control (see Supplemental Table 3 for details). The study endpoints included pain VAS, NDI, and credibility and expectancy questionnaire. The investigators did report benefit in medium or high force interventions.

Additional information on all neck pain RCT’s can be found in Supplemental Table 3.

OSTEOARTHRITIS OF THE KNEE

Acupuncture

Four RCT’s examined whether acupuncture needling could relieve symptoms associated with osteoarthritis of the knee.^87–90 These studies used similar definitions of knee OA. Participants were predominately female; had mean ages between 60 and 65; and had knee pain for an average of 9 to 11 years. All studies incorporated either the Western Ontario and McMaster Universities (WOMAC) arthritis Index total score or the WOMAC pain subscore as the primary outcomes. In two of the trials,^{89, 90} no difference was seen between verum acupuncture and sham acupuncture for either the primary or secondary outcome measure. The third trial⁸⁷ using a sham control saw significantly better improvement in both the WOMAC pain subscale and function subscale after 14 weeks of treatment. For the three trials that also incorporated either an attention control⁸⁷ or standard care controls,^{88, 90} verum acupuncture produced significantly better improvement in primary and secondary outcomes than did the control group. Across trials, adverse events associated with acupuncture were few in number with the most common complaints being pain at the needling site and muscle soreness that resolved soon after the treatment session ended.

Glucosamine and Chondroitin

We reviewed eight RCT’s examining the efficacy of two dietary supplements, glucosamine and/or chondroitin, in individuals with knee OA.^91–98 These studies varied substantially in how knee OA was defined, as well as in the primary outcome measures used (see Supplemental Table 4).

Three different configurations of glucosamine were used: glucosamine hydrochloride (HCl),^93–95 glucosamine sulfate,^{91, 92, 98} and glucosamine bound to a polymer.⁹⁷ One trial did not identify the configuration of glucosamine.⁹⁶ Doses of glucosamine varied from 1000 mg per day for 6 weeks⁹⁸ to 1500 mg per day for 24 weeks.^{91, 95} Chondroitin sulfate was studied in two trials at a dose 1200 mg/day for 6 months^{91, 95} and in one trial at the same dose but for 8 weeks.⁹³

All studies employed a placebo control group. Participants in two trials were made up exclusively⁹³ or predominately of males;⁹⁶ participants in four other trials were predominately females;^{91, 94, 95, 98} the remaining two trial were closely balanced in men and women.^{92, 97} Mean participant ages for the trials varied from 45 years of age⁹³ to 72 years of age.⁹⁵

Mixed results were found in trials comparing glucosamine versus a placebo control with pain relief or functional improvement as the primary outcomes. Three trials found glucosamine superior to placebo,^{92, 97, 98} and three trials failed to see a difference between groups.^{91, 94, 96}

Mixed results were seen in the three trials studying a combination of glucosamine and chondroitin. Leffler et al.⁹³ found that a combination of glucosamine HCl and chondroitin sulfate (and manganese acerbate) was superior to placebo for providing pain relief. In two other trials, the combination of glucosamine (either sulfate or HCl) and chondroitin sulfate were no better than placebo for either pain relief or function.^{91, 95}

Across trials, adverse events were generally mild (gastrointestinal distress being the most common) with no differences seen between those taking glucosamine or chondroitin and those taking placebo.

Massage therapy

Two RCT’s studied the efficacy of Swedish massage therapy for symptoms associated with osteoarthritis of the knee.^{99, 100} The two studies were similar in that both: 1) defined knee OA using the American College of Rheumatology criteria;¹⁰¹ 2) required a baseline score of at least 40 on the visual analog pain scale; 3) participants were predominately white females. In the earlier study¹⁰⁰, after completing the 8 week intervention, participants in the massage group saw significant improvement versus those randomized to Usual Care in the WOMAC Arthritis Index, as well as in each of the WOMAC subscales: pain, function and joint stiffness. In a dosing study,⁹⁹ individuals were randomized to one of five groups for 8 weeks: 1) Usual care; 2) 240 minutes of massage over 8 weeks; 3) 360 minutes of massage; 4) 480 minutes of massage; and 5) 600 minutes of massage. Only individuals receiving at least 480 minutes of massage therapy (groups 4 and 5), saw substantial improvement in the WOMAC OA index and the WOMAC Pain subscale versus the Usual Care group. Across the two trials only one adverse event, discomfort at the knee in one participant, was noted.

Methylsulfonylmethane (MSM)

One RCT compared MSM (6 grams/day for 12 weeks) to a placebo control.¹⁰² Knee OA was based on ACR criteria.¹⁰¹ Outcome measures were the WOMAC arthritis subscales for pain, function, and stiffness. Individuals randomized to MSM saw significant improvement in both the WOMAC pain and functions scales. However, the authors cautioned that the differences between groups were small (<20%) and probably not clinical meaningful. No difference in adverse events was seen between groups.

S-Adenoysl methionine (SAMe)

One RCT compared the dietary supplement SAMe (1200 mg/day) to celecoxib (200 mg/day for 16 weeks in a cross-over design that included a one-week washout period.¹⁰³ Knee OA was defined based on ACR criteria:¹⁰² the population was primarily female and white. The primary outcomes were VAS pain scores, WOMAC arthritis subscales for pain, function, and stiffness, and clinician assessments of OA severity. By the end of the trial, no differences were seen between the two treatment arms on any of the outcomes measures. However, the sample size was insufficient to establish equivalency. Overall adverse events were less common in those taking SAMe then those taking celecoxib.

Tai Chi

We reviewed four RCT’s that examined the efficacy of Tai Chi in those with confirmed knee osteoarthritis using various criteria for diagnosis.^104–107 The four RCT’s had similar study populations, primarily white and female, with mean ages ranging from 65 to 79. Mean BMI was also similar across studied ranging from 27.8 to 30.

In Hartman et al.,¹⁰⁵ participants randomized to a Yang style tai chi group had significant improvement in the Arthritis Self-Efficacy scale versus those in the Routine Care group. The remaining trials compared tai chi to attention controls.^{104, 106, 107} All found that Tai Chi (either Yang style or Sun Style) did better than the attention control group on both primary and secondary outcomes. Across trials the most common adverse event associated with Tai Chi was minor muscle soreness that resolved after a few days.

Yoga

Park et al.¹⁰⁸ completed a quasi-experimental trial, with participants randomized to either sitting yoga or Reiki; however the attention control group was chosen as a convenience sample from other participants meeting the inclusion/exclusion criteria. The primary outcome measures were the WOMAC Arthritis Index score and the pain, function, and stiffness subscales. The mean age of participants was 80, with 68.7% being male. Individuals randomized to yoga, had substantially better reduction in the WOMAC function then either the Reiki group or the attention control group (p<0.02). No other differences were seen between the three groups. No adverse advents were reported for any of the groups in this study.

Additional information on all osteoarthritis RCT’s can be found in Supplemental Table 4.

SEVERE HEADACHE AND MIGRAINE

The majority of studies we reviewed defined cervicogenic headache, migraine, and tension-type headache with the International Classification of Headache Disorders second edition.^{109, 110}

Acupuncture

Coeytaux et al.¹¹¹ randomized patients with chronic daily headache to management by a neurologist with or without acupuncture. The acupuncture group had significantly reduced headache impact test scores versus the control, but there were no significant differences on pain severity.

Massage

One small study that compared myofascial trigger point massage, a sham device, and waitlist; found massage reduced headache frequency but did not significantly reduce intensity or duration of headaches.¹¹²

Omega-3 Fatty Acids

One study randomized adolescents with migraine to omega-3 or placebo in a cross-over study.¹¹³ Adolescents experienced a reduction in headache frequency when taking either fish oil or placebo, but there was no significant difference between the treatments. Another study randomized patients with chronic headaches to increased omega-3 and lower omega-6 fatty acids in the diet or to lower omega-6 fatty acids in the diet.¹¹⁴ The participants on high omega-3 and low omega-6 diet had greater improvement on the Headache Impact Test and number of headache days.

Relaxation techniques

Six trials we reviewed evaluated stress management, relaxation, or biofeedback for headache disorders.^115–120 Slavin-Spennyt et al.¹²⁰ randomized students with chronic headache to expression training, relaxation training, or waitlist. Both active treatments showed significant decreases in headache frequency versus the waitlist control, but did not differ among themselves. Two trials examined complementary approaches to either tension headaches or migraine.^{116, 117} D’Souza et al. (2008)¹¹⁷ randomized students to relaxation training, written emotional disclosure, or neutral writing. The relaxation group had greater reductions in headache frequency and the associated headache disability compared to the other two groups. Devineni et al.¹¹⁶ randomized participants to an internet behavioral intervention or a waitlist. The behavioral intervention demonstrated improvement in the headache index versus waitlist. A trial randomized children with migraine to hand warming biofeedback with stress management, hand cooling biofeedback (attention control), or waitlist.¹¹⁹ The warming biofeedback group had improved headache index scores compared to the other groups. Holroyd et al.¹¹⁸ randomized participants with chronic tension headache to tricyclic antidepressant, placebo, stress management, or stress management plus tricyclic antidepressant. Both the medication and stress management groups had improvements over placebo, but the combination produced the best outcomes. Blanchard et al.¹¹⁵ randomized patients with headache to biofeedback with relaxation training, biofeedback plus cognitive therapy, sham meditation, or a headache monitoring control condition. All of the treatment groups including the “sham” meditation group had improvements in the headache index in comparison to the monitoring control group.

Spinal and Osteopathic Manipulation

One trial randomized patients with chronic cervicogenic headache to two doses of spinal manipulation or light massage;¹²¹ based on the literature, the light massage should have little, if any, specific effects and was therefore chosen as a control for time and physical contract with the patient. Haas and colleagues found improvement over all time points favoring spinal manipulation compared to light massage and a dose effect with the spinal manipulation. Two RCT’s were identified that assessed manipulation in individuals with tension headache.^{122, 123} The first trial randomized patients with chronic tension type headache to spinal manipulation or amitriptyline medication and found no differences between the groups;¹²² however the trial did not appear powered to detect non-inferiority. The second study was a small trial that randomized patients to osteopathic manipulation, a palpation examination, or no treatment.¹²³ Authors noted an improvement in headache severity for the spinal manipulation group; however no statistical comparisons were made between groups. Nelson and colleagues study randomized patients with migraine to amitriptyline, spinal manipulation, or both treatments.¹²⁴ The study found no significant differences between the groups; however it did not appear that the trial was powered to detect non-inferiority (see Supplemental Table 5).

Additional information on all headache RCT’s can be found in Supplemental Table 5.

OVERALL SUMMARY OF RCT DATA

Tables 3 (Efficacy), and 4 (Effectiveness) provide concise summaries of the reviewed clinical trial data for each complementary approach across painful health conditions. In these tables, positive trials are those in which the complementary approach provided statistically significant improvements in pain severity or pain-related disability or function compared to the control group. Negative trials are those in which no difference was seen between groups. Based on a preponderance of positive trials versus negative trials, current evidence suggests that the following complementary approaches may help some patients manage their painful health conditions:

Table 3.

Summary of Evidence for Selected Complementary Health Approaches by Type of Pain - sham or placebo, attention controls

Approach	Back pain	Fibromyalgia	OAa of knee	Neck pain	Severe Headache/Migraine
Acupuncture	1 positive trial, 2 negative	1 positive trial, 3 negative trials	1 positive trial, 3 negative	?b	? b
Chondroitin	? b	? b	1 negative trial	? b	? b
Glucosamine	? b	? b	2 positive trials, 3 negative trials	? b	?
Chondroitin & Glucosamine		?	1 positive trial, 2 negative trials	?	? b
Massage therapy	One positive trial	? b	? b	2 positive trials	One positive trial
MSMa	? b	? b	1 positive trial	? b	? b
Omega 3 fatty acids	? b	? b	? b	? b	One negative trial
Relaxation approaches	? b	1 positive trial, 2 negative	?	? b	3 positive trials
SAMea	? b	? b		? b	? b
Spinal Manipulation	6 positive trials, 3 negative	? b	? b	One negative trial	One positive trial
Osteopathic Manipulation	1 positive trial, 1 negative	? b	? b	? b	? b
Tai Chi	? b	2 positive trials	3 positive trials	? b	?
Yoga	One positive trial	b?	1 positive trial	? b	? b

^aAbbreviations: MSM = Methylsulfonylmethane; OA = osteoarthritis; SAMe = S-Adenoysl methionine

^bno U.S. RCT’s identified

• Acupuncture and Yoga for back pain;

• Acupuncture and Tai Chi for osteoarthritis of the knee;

• Massage therapy for neck pain with adequate doses and for short term benefit; and

• Relaxation techniques for severe headaches and migraine.

Weaker evidence suggests that Massage therapy, Spinal Manipulation and Osteopathic Manipulation might also be of some benefit to those with Back Pain, while Relaxation approaches and Tai Chi might help those with Fibromyalgia.

SAFETY

Generally the reporting of safety data in the reviewed RCTs was minimal. For those trials that did report safety data, we have summarized this information in the text for each painful health condition. In no case did an RCT identify a serious adverse event associated with any of the complementary approaches examined. The most common adverse events (gastrointestinal distress) were noted in trials of dietary supplements (glucosamine, chondroitin, MSM, SAMe). In some trials, tai chi and yoga were associated with minor muscle or joint soreness, and acupuncture was associated with minor pain and/or bruising at the needling site.

Comparisons to recent systematic reviews

Our search criteria identified a number of recent (2010 or later) systematic reviews that covered our topics of interest.^125–133 Conclusions from these systematic reviews for practitioner approaches (acupuncture, chiropractic, massage therapy)^125–131 and dietary supplements¹³² were generally consistent with our findings. For instance, in a comprehensive reviews of both pharmacological and non-pharmacological approaches to manage back pain, Chou et al. (20016)¹²⁵ found that acupuncture and yoga appear to be effective for improving pain and/or function in back pain. A Cochrane systematic review concluded that acupuncture was a viable treatment option for osteoarthritis of the knee.¹²⁷ Also concurring with the present analysis, the meta-analysis by Deare et al.¹²⁶ concluded that acupuncture was not an effective therapy for pain or function in individuals with Fibromyalgia. In their systematic review, Posadzki and Ernst ¹²⁸ found little data supporting the use of spinal manipulation for headaches. Supporting our conclusions is a recent comprehensive meta-analysis of trials studying glucosamine for osteoarthritis,¹³² which found considerable variability in results across trials and concluded that neither glucosamine sulfate or HCl provide pain relief. The RCT’s examined in two systematic reviews of yoga for arthritis^129,130 overlapped considerably with RCTs in the present review. The authors of the systematic reviews concluded that yoga appears to be a viable option for relieving pain and discomfort associated with arthritis, but that larger, better designed trials were needed. A recent systematic review¹³¹ that included international trials found that “clinically relevant effects of OMT were found for reducing pain and improving functional status” for those with back pain. Although this is a stronger conclusion than we reached, the authors’ identified deficiencies in trial methodology and called for larger, better quality RCT’s to provide firm conclusions. Not all recent systematic reviews agreed with our conclusions. A recent Cochrane Collaboration meta-analysis¹³³ concluded that spinal manipulation was no more effective than “inert” interventions for managing back pain and related disability. However this review only included RCT’s published through 2009. We reviewed eight RCTs published since then. Of these, two were negative trials^32,27 and six were positive trials.^{36,41–43,46,54} Inclusion of these trials into the meta-analysis might have lead Rubenstein et al.¹³³ to draw a different conclusion.

Caveats

A number of methodological issues temper our conclusions: 1) The trial samples tend to be white, female and older, with very few, if any minority group participants; as such the generalizability of the findings to the breath of patients seen by primary care providers in the U.S. is still unresolved; 2) Often the trials reviewed were small, with fewer than 100 total participants. Small trials are prone to more variability and to false-negative results. 3) In many of the trials where the statistical superiority of a given complementary health approach was demonstrated, it was not clear if the differences versus the control group were clinically relevant; 4) For the given painful health condition, a wide number of outcome measures were often used to asses pain and function; this plethora of outcomes may partly explain the conflicting results seen across trials; 5) For most complementary approaches there are no standard treatment protocols or algorithms, or in the case of dietary supplements, no rigorously established dosages and products; as such, trials of a given complementary approach rarely compare the exact same intervention; and 6) Our findings that relatively few mild adverse events and no serious adverse events were associated with complementary approaches is consistent the findings from a number of systematic reviews.^125–133 However, even large clinical trials are not powered to identify infrequent adverse events; therefore, it is likely that this review is underestimating the entire range of events associated with the complementary approaches examined. Finally, our review was intended to be an overview of data from RCTs performed in the US. The inclusion of RCTs performed outside the US may have resulted in a different set of recommendations.

Table 4.

Summary of Evidence for Selected Complementary Health Approaches by Type of Pain - wait list, usual care, or routine care controls

Approach	Back pain	Fibromyalgia	OAa of knee	Neck pain	Severe Headache/Migraine
Acupuncture	2 positive trials	?b	2 positive trials	? b	One positive trial
Massage therapy	3 positive trials, one negative	? b	2 positive trials	3 positive trials	? b
Natural products supplements	? b	? b	? b	? b	? b
Relaxation approaches	? b	4 positive trials	? b	? b	4 positive trials
Spinal Manipulation	4 positive trials, 3 negative	? b	? b	? b	? b
Osteopathic Manipulation	2 positive trials, 2 negative	? b	? b	? b	? b
Tai Chi	? b	? b	1 positive trial	? b	? b
Yoga	4 positive trials	One positive trial	? b	? b	? b

^a= Abbreviations: OA = osteoarthritis

^b= no U.S. RCT’s identified

List of abbreviations

ACR

American College of Rheumatology

cLBP

chronic low back pain

EMS

electrical muscle stimulation

FIQ

Fibromyalgia Impact Questionnaire

HVLA-SM

high velocity, low amplitude, spinal manipulation

LBP

low back pain

MSM

Methylsulfonylmethane

NDI

Neck Disability Index

NRS

numeric rating scale

OA

osteoarthritis

ODI

Oswestry Disability index

OMT

osteopathic manipulative therapy

RCT

randomized clinical trial

RMDQ

Roland-Morris Disability Questionnaire

ROM

range of motion

SAMe

S-Adenoysl methionine

SM

spinal manipulation

VAS

visual analog scale

WOMAC

Western Ontario and McMaster Universities Arthritis Index