Published Trials of Nonmedicinal and Noninvasive Therapies for Hip and Knee Osteoarthritis

Methods

We reviewed all published trials that 1) included patients with osteoarthritis of the hip or knee; 2) assigned patients to a treatment group consisting of nonmedicinal, noninvasive therapy or to a concurrent control group that received either no therapy or an alternate or placebo intervention; and 3) included pain or a measure of lower extremity function as an outcome. Articles of interest were identified using MEDLINE by cross-referencing the medical therapy subheadings of “osteoarthritis,” including “drug therapy,” “radiotherapy,” “surgery,” and “rehabilitation,” with “clinical trials,” “comparative study,” or “trial(s)” from 1966 through 1993. Of the 600 English-language articles on humans retrieved, 452 addressed medical therapy, of which 377 (63% of the total) involved nonsteroidal anti-inflammatory drugs. In contrast, only 31 reports dealt with nonmedicinal, noninvasive therapies. Of these, 11 met the three inclusion criteria previously listed [8-18], and bibliographies of these articles provided 3 additional studies [19-21]. The 14 articles (15 trials) are summarized in Table 1. The remaining 20 reports were studies of osteoarthritis of other anatomic sites (n = 8), were nonexperimental studies, or did not include pain or function as an outcome measure (n = 12).

In evaluating the quality of the trials, we used criteria developed by Sackett and colleagues [22] to analyze therapeutic efficacy, including whether assignment was random, clinically relevant outcomes were reported, patients had symptoms typical of the disease studied, clinical and statistical significance was considered, the therapeutic maneuver was subject to contamination or cointervention, and whether follow-up was complete. We also considered whether assessment was done by a blinded observer and whether the participants were blinded to treatment. For many of the treatments, blinding of the study participants was difficult (with exercise trials, for example). Sham treatments designed to create double-blinding for ultrasound, laser, transcutaneous electrical nerve stimulation, and pulse electromagnetic fields consisted of similar devices that appeared to the patient to be “on” but did not transmit the associated therapy. For acupuncture, sham treatment consisted of placing needles at nontraditional acupuncture points. We judged the adequacy of sample size using guidelines developed by Bellamy and colleagues [23-25] for the number of participants needed to detect clinically meaningful and statistically significant differences for frequently used outcomes (when performed in a standardized manner) in osteoarthritis antirheumatic drug trials.

The most common instrument for pain assessment in these trials was self-reported pain using a visual analog scale (a numeric scale with 0 indicating no pain and 10, the worst pain); the outcome measure reported was change in pain from baseline. The absolute change in score was usually divided by the entire range of the scale or the baseline value. In reporting improvement in pain in these studies, we used the former when the data were available to do so. The most common objective function measurement of the lower extremity was a timed walk; change was reported as a percentage change from baseline.

Results

Diathermy

Diathermy (“heating through” or deep heat) can be delivered by high-frequency sound waves (ultrasound) or electromechanical irradiation (microwave or short wave). Although sometimes prescribed for pain control alone, deep heat is typically applied before exercise involving large joints with limitations of joint motion to facilitate tendon extensibility and muscle relaxation. We identified two randomized, controlled trials of diathermy [8, 12].

Chamberlain and colleagues [12] randomly assigned 42 patients with well-defined osteoarthritis of the knee to receive short-wave diathermy applied to affected joints three times weekly by trained personnel, followed by exercises plus instructions regarding additional twice-a-day home exercise (n = 24), or similar exercise instruction only (n = 18). The exercises prescribed were designed to strengthen the knee extensors by graded resistive isotonic exercise (requires joint motion). At 4 weeks, 21 (88%) patients receiving diathermy plus exercise and 15 (83%) patients receiving exercise therapy alone were evaluated by a research assistant blinded to treatment arm. Both groups reported a reduction in pain measured on a visual analog scale and improved function based on a summated score derived from the ability to perform several activities, including walking and stair climbing. However, the only between-group difference was greater relief of pain (P < 0.05) in the exercise-only group. The data given were insufficient to determine whether the difference in pain was clinically meaningful or whether the sample size or scale used was sufficient to detect meaningful differences in pain or function. The exercise regimen was not standardized, and thus differences in exercise performed rather than the presence of diathermy could have caused the observed difference in pain relief.

Falconer and colleagues [8] randomly assigned 74 patients with well-defined knee osteoarthritis associated with limited knee motion lasting for at least 6 months to receive ultrasound followed by a 30-minute exercise protocol for stretching and strengthening done 2 to 3 times a week for 4 to 6 weeks or an identical program that used sham ultrasound. When the program was completed, an experienced physical therapist blinded to treatment arm evaluated 95% and 92% of experimental and control participants, respectively. Study participants had a 20% decrease in pain (visual analog scale), a 12% increase in gait velocity (time to walk 50 feet), and a 9% increase in active knee range of motion; however, no between-group differences occurred (P > 0.05). The study was designed to detect a 5.3-degree difference in change of knee range of motion between groups and was also sufficient to detect clinically significant differences in pain and gait velocity [25]. Ultrasound is usually done before exercise, so the exercise cointervention was appropriate; in this study, the exercise was supervised and therefore should have been similar in the two groups.

Exercise

Exercise refers to a broad range of activities involving active or passive joint movement and muscle contraction. Many studies have documented joint motion restriction, muscle weakness, and deconditioning in patients with degenerative arthritis [26-28]. In osteoarthritis, muscle rehabilitation correlates with improved functional and psychological status [29].

Quadriceps Strengthening

Two controlled trials evaluated quadriceps strengthening in patients with knee osteoarthritis. Jan and Lai [9] alternately assigned 61 women with well-defined knee osteoarthritis to receive diathermy (random assignment to receive ultrasound or short-wave diathermy) alone or diathermy plus exercise. The exercise consisted of isometric (no joint motion) knee extensor strengthening by 200 straight-leg raises daily. Diathermy was done at least 4 times per week, and the length of follow-up varied. A “functional incapacity score” was calculated from patient reports of pain and disability during various daily activities (modified Bandi criteria), and muscle strength was measured using an isokinetic dynamometer in an isometric testing mode. The exercise group showed a significantly greater improvement in the incapacity score compared with the nonexercise control group (30% compared with 20%; P < 0.01) and also showed greater improvement in measured muscle strength (P < 0.01). No attempt was made to determine compliance with the exercise regimen, and thus we do not know whether participants actually did the 200 leg raises daily. Also, the length of follow-up varied, was not well described, and may have varied between treatment groups.

Chamberlain and colleagues [12] re-randomized participants who completed their exercise and diathermy trial (noted previously) at 6 weeks on the basis of whether they were informed about plans for assessment again at 12 weeks or not informed; 29 of 36 (81%) participants completed this second evaluation. They had been instructed to perform twice-daily home exercises designed to strengthen the knee extensors by a graded resistive isotonic exercise (which requires joint motion) program using weights to be increased by 0.5 pounds at intervals as tolerated up to a maximum of 10 pounds. Exercises were to be done “within the limits of pain and muscle fatigue.” Ten of the 14 participants in the informed group (“exercise group”) reported that they continued the prescribed exercises, in contrast to only 2 of the 15 uninformed patients (“control group”). Pain (visual analog scale) and function (summated scale) scores were similar at 6 weeks (time of re-randomization) but were significantly better (pain, P < 0.001; function, P < 0.05) in the exercise group at 12 weeks. We could not quantify the degree of improvement in either scale from the data presented. In addition, we found no details on the exercise (length of time, number of sessions, use of weights) actually performed.

Aerobic Activity

Often aerobic exercise is combined with range-of-motion stretching and strengthening exercise. Two controlled trials evaluated the efficacy of aerobic exercise combined with stretching and strengthening in knee or hip osteoarthritis.

Minor and colleagues [13] randomly assigned 120 patients with well-defined rheumatoid arthritis (n = 40) or osteoarthritis (n = 80) of the hip, knee, or ankle to receive one of three interventions: a stretching and strengthening exercise program alone, or the same program combined with one of two aerobic exercises: pool activities or walking. All patients participated in a 12-week program, which met 3 times each week for 1 hour, and all performed supervised range-of-motion and isometric exercises. The two aerobic exercise groups also did up to 30 minutes of exercises to increase their heart rates to 60% to 80% of each person's baseline maximum by walking or pool activity. At 12 weeks, 80% of study participants were assessed for aerobic capacity (maximum oxygen uptake), flexibility (trunk bending), function tests (exercise endurance on a treadmill test, time to walk 50 feet, and reported daily activity), and self-reported health status using the Arthritis Impact Measurement Scale (AIMS), an arthritis-specific functional status instrument that reliably measures psychological health, physical health, and pain [30]. The pool and walking programs increased aerobic capacity 20% and 19%, respectively, compared with no change in the control group (P = 0.009). The AIMS scores based on physical activity (P = 0.009), anxiety (P = 0.004), and depression (P = 0.007) subscales improved from 5% to 12% in both aerobic groups, significantly more than for the control group. There were no significant intergroup differences in the pain subscale (6%, 12%, and 7% reduction for the visual analog scale in the pool, walk, and control groups, respectively). Both aerobic groups decreased their times to walk 50 feet by 12%, compared with a 2% change for control participants (P = 0.009). Although trends were similar in a subgroup analysis evaluating osteoarthritis alone, the study was not designed to analyze osteoarthritis alone and had inadequate power to compare the aerobic and nonaerobic (control) exercise groups. The cointervention (stretching and strengthening) was identical in all three groups. The articles contained no mention of masking the assessors in this trial.

Kovar and colleagues [10] randomly assigned 102 patients with documented primary knee osteoarthritis to either an 8-week program that included lectures, group discussions, and supervised light stretching and strengthening followed by as much as 30 minutes of walking or to routine care and telephone follow-up three times per week. Evaluation at baseline and 8 weeks by a physical therapist who was not blinded to treatment group was complete for 47 of 51 (92%) patients receiving the intervention and 45 of 51 (88%) controls. Intervention participants had a 70-meter increase (18%) in 6-minute walking time compared with a 17-meter decrease in controls (P < 0.001). The intervention group also improved more than controls on the self-reported AIMS subscales for physical activity (24% compared with −2%; P < 0.001), arthritis impact (17% compared with 8%; P = 0.09), arthritis pain (14% compared with 1%; P = 0.003), and use of medications (26% compared with 4%; P = 0.08). The independent effects of aerobic exercise and the other cointerventions in this study (stretching and strengthening, patient education, and intensity of attention) could not be determined.

Topical Irritants

Patients may apply topical irritants to painful peripheral joints and muscles, but this type of therapy is impractical for hip disease. We identified only one trial of such therapy for knee osteoarthritis, and it involved capsaicin (an extract from chili peppers found in Tabasco sauce). Although capsaicin produces a local burning sensation, it may also interfere with the production of substance P, a chemical mediator of pain from local sensory terminals to the central nervous system [31].

In a randomized, double-blind, placebo-controlled multicenter trial involving 70 patients with knee osteoarthritis and moderate to very severe pain, Deal and colleagues [16] instructed patients to apply 0.025% capsaicin cream or a placebo four times daily to the front, back, and both sides of their most severely affected knee. Thirty-four of 36 (94%) patients using active therapy and 30 of 34 (88%) using a placebo were evaluated at 1, 2, and 4 weeks. Those using active treatment improved significantly more using the physician's global (5-point scale) evaluation of pain (P = 0.02) and the patient's (visual analog scale) pain score (P = 0.03) than did those using placebo for the three post-baseline evaluations. At 4 weeks, a 22% reduction occurred in pain by physician evaluation and patient score in those using the active cream compared with 14% (P = 0.05) and 10% (P = 0.06) in those using placebo cream. Burning was noted by 44% of patients with initial application of capsaicin, but it decreased or vanished with repeated use; this property made complete blinding of study treatment impossible.

We found no other trials of capsaicin or other topical creams (excluding those containing nonsteroidal anti-inflammatory drugs) for hip or knee osteoarthritis.

Laser

Low-energy laser therapy has been used to control pain with many diseases, but its mechanism of action is not known. The energy delivered is too low to produce heat, but some data suggest that nerve transmission may be altered [32].

Stelian and colleagues [17] randomly assigned 50 patients with well-defined osteoarthritis of both knees associated with at least 3 months of pain to receive treatment with red (wavelength, 633 nm), infrared (830 nm), or placebo laser light emitters. The placebo emitter was similar to the infrared but not the red emitter, and thus the comparison between placebo and infrared therapy was double-blind. Treatment was applied at both sides of the knee twice daily for 10 days; measurements were available for all 50 participants at baseline and after 10 days of treatment. Patients in the two active treatment groups had significantly greater pain reduction than did those in the placebo group (P < 0.05) using three scales: a visual analog scale, the Short-form McGill Pain Questionnaire, and the Present Pain Intensity scale. Patients receiving red and infrared light treatment had improved pain scales of 32% and 39%, respectively (P = 0.00001 when compared with baseline for both), but participants in the placebo group showed no change (0.6% increase; P = 0.8). Patients in the red and infrared treatment groups also had greater decreases in self-reported functional disability, 40% (P = 0.007) and 49% (P = 0.0001), respectively, than did the placebo recipients (7%; P = 0.2) using the Disability Index Questionnaire, which measures difficulty in performing nine common activities. No objective measurement of lower extremity function was done. Pain relief was long-lasting in many patients, and patients requested retreatment because of recurrent pain at an average of 4.2 and 6.1 months in the red and infrared treatment groups compared with 0.5 months in the placebo group (P < 0.05). The degree of patient improvement, the duration of benefit, and the small placebo effect reported are atypical compared with other osteoarthritis trials.

Walker [21] randomly assigned 36 patients with various chronic pain syndromes (including 8 with osteoarthritis) to receive irradiation with a helium-neon laser (632 nm) designed to achieve analgesia at the painful areas (n = 26) or control sites (n = 10). Active treatment involved 3 weekly sessions for 10 weeks, during which the radial, median, ulnar, and saphenous nerves received 20 seconds of irradiation and the painful site received 30 seconds, with a weekly increase at this site as tolerated. Control participants received radiation using the same schedule to selected other areas and to a skin site adjacent to the painful area. In the intervention group, 19 of 26 patients (including 4 of 5 with osteoarthritis) noted more than 50% pain reduction compared with 1 of 10 control participants (including 0 of 3 with osteoarthritis). The number of patients with osteoarthritis was too small to detect meaningful differences, and the heterogeneous diagnoses made interpretation of these results difficult.

Acupuncture

Acupuncture has been used to treat various illnesses for more than 2000 years and has been applied to patients with osteoarthritis. Two controlled trials of acupuncture that met our inclusion criteria were reviewed.

Gaw and colleagues [11] conducted a double-blind trial in 40 patients with osteoarthritis and pain affecting the neck, hands, or weight-bearing joints who were randomly assigned to receive acupuncture with needles applied at traditional sites or sham acupuncture with needles placed at nontraditional “placebo points.” Treatment was done weekly for 8 weeks and consisted of three 5-minute periods of manual twirling of inserted needles, with intervals of rest, without removing the needles. Two physicians independently assessed 38 of 40 (95%) patients' responses to treatment, including physician-determined joint tenderness and patients' self-reported pain and activity at baseline and within 36 hours of treatment using numeric scales from 1 to 4. After 8 weeks of therapy, modest improvement occurred in all three scales (about 5% to 12%), but no intergroup differences were noted (P > 0.05). The correlation of the two independent assessors of self-reported pain, the major clinically significant outcome, was fairly good (Spearman rank-order correlation coefficient, 0.67), and the sample size was adequate to detect clinically significant between-group differences in this measure [25]. Although all patients improved with treatment, none of the improvement could be attributed to traditional acupuncture. Explanations for the improvement include natural history of the disease, some other aspect of trial participation, or a true effect of needle treatments, regardless of their placement at traditional or nontraditional acupuncture sites. Because the study included patients with osteoarthritis of other joints besides the hip and knee, the efficacy of acupuncture in some joints could have been masked by inefficacy in others.

Christensen and colleagues [19] studied the effects of acupuncture on pain and function in patients with osteoarthritis of the knee who were on a waiting list to receive total knee arthroplasty. Thirty-two patients were randomly assigned to receive 20 minutes of acupuncture with five needles placed around the painful knee and one in the ipsi-lateral hand twice weekly for 3 weeks or no acupuncture; 29 (91%) had baseline and 9-week evaluations. At week 9, the treated group reported about a 23% decrease in pain (median value using the visual analog scale) compared with a 12% increase in the untreated group (P = 0.03), and the median time to walk 50 meters and 20 steps decreased 28% and 30%, respectively, compared with a decrease of 13% and increase of 9%, respectively, in controls (P < 0.01 for both comparisons). Subsequent treatment of the control group resulted in score changes similar to those reported by the intervention group. No sham treatment was used, so patients' assessments and performances could have been influenced by placebo effect or the visits with the therapist rather than by the treatment itself. However, objective assessments of performance times were made by a blinded observer.

Transcutaneous Nerve Stimulation

Transcutaneous electrical nerve stimulation therapy has been used to treat many acute and chronic painful conditions. Three randomized, controlled trials were done in patients with osteoarthritis of the knee. Lewis and colleagues [14] randomly assigned 30 patients with at least 12 months of pain attributed to osteoarthritis of the knee to receive active transcutaneous electrical nerve stimulation and placebo treatment in a double-blind cross-over trial. Patients self-administered active or sham transcutaneous electrical nerve stimulation 3 times daily for 30 to 60 minutes for 3 weeks. Of the 28 (93%) who completed 3 weeks of both therapies, 46% reported more than 50% pain relief with active therapy compared with 43% with sham therapy (P > 0.05 for between-group difference). The absolute change in pain score (visual analog scale) was similar during active (32%) and sham therapy (25%, P > 0.05), but the duration of benefit was significantly longer after active treatment (151 compared with 110 minutes; P < 0.01). In addition, 12 of the 28 patients asked to continue active therapy after the trial, whereas 4 asked to continue placebo treatment. Function was not measured and long-term effects were not studied. The report did not state whether the assessors were blinded to treatment arm.

Taylor and colleagues [15] completed a 2-week randomized, double-blind, cross-over trial comparing active and placebo transcutaneous electrical nerve stimulation in 10 of 12 (83%) patients with symptomatic knee osteoarthritis who were candidates for total knee replacement. Patients were instructed in self-administration of the device for periods of 30 to 60 minutes on an as-needed basis, and, at the end of 2 weeks, they were interviewed about change from baseline in pain relief (using two 5-point scales), walking ability (4-point scale), and medication use (4-point scale). Patients reported significantly more pain relief (P = 0.03) using one of the two pain scales and less medication use (P = 0.06) with active compared with placebo therapy. Most patients reported pain relief only while actively using the device, although a few patients noted a sustained effect for several hours. As in the previous study, 50% of patients (5 of 10) asked for active therapy after the trial, and only 1 patient requested placebo. Two of the five patients still received active therapy 1 year after the study. The small sample size did not permit detection of changes in walking ability [25].

Fargas-Babjak and colleagues [20] studied the effects of an acupuncture-like transcutaneous electrical nerve stimulation machine (Codetron) in which six electrodes were placed at predetermined sites (depending on the joint involved) and stimulated randomly. Fifty-six patients with at least 6 months of symptoms related to hip or knee osteoarthritis were randomly assigned to receive genuine or sham treatment and instructed in home use of the devices for 30 minutes twice daily. Patients who could not or did not comply were excluded, leaving 37 (66%) patients with follow-up at 6 weeks. Of these, 74% of patients using active treatment reported more than a 25% reduction in pain (visual analog scale) compared with 25% of controls (P < 0.02). No differences were observed in the time to walk 50 feet, but, again, the sample size was too small [25].

Pulsed Electromagnetic Fields

Pulsed electromagnetic fields have been used to treat delayed union fractures for more than 10 years. A system that delivers low-frequency pulsed waves was developed to treat patients with various painful rheumatic conditions [18]. One of the developers of this system and colleagues [18] tested this therapy in a trial that included 27 patients with persistent symptomatic osteoarthritis (21 patients with knee osteoarthritis). Patients were randomly assigned to receive treatment with pulsed electromagnetic fields for 30 minutes, 3 to 5 times per week, for a total of 18 treatments per month or identical sham treatment in a double-blind manner. Fourteen of 15 (93%) patients assigned to receive active therapy and 10 of 11 (91%) assigned to receive sham treatment had baseline and post-treatment evaluation of overall and worst pain (visual analog scale), difficulty and pain with activities of daily living (scale of 1 to 5), and pain and joint tenderness by physician examination (scale of 1 to 5). Patients assigned to receive active treatment had a 39% reduction in overall pain (P = 0.002) compared with a 8% reduction (P > 0.6) in the control group. Changes in the other five scales were similar to those reported for overall pain for both active and control groups. No objective measurement of function was made.

Other Therapies

Application of heat and cold has long been recommended for short-term pain relief associated with many musculoskeletal conditions. Superficial heat can be delivered by moist or dry packs, warm baths including spas, and infrared light; ice packs or vapocoolant sprays provide topical cooling. Superficial heat is commonly prescribed before stretching exercise to increase collagen elasticity and decrease muscle spasm, whereas cold is used after strenuous exercise to relieve muscle soreness. Thermal therapy has considerable anecdotal and historical precedent; in fact, the original rheumatologic hospitals arose around natural hot springs [33]. In a recent survey of patients with rheumatoid arthritis and osteoarthritis who were attending a rheumatology clinic, approximately 60% reported using topically applied heat or warm baths, and 22% applied cold to painful areas [34]. We found no controlled trials of superficial heat or cold application for osteoarthritis.

Orthoses and assistive devices limit weight bearing, decrease motion around a painful joint, and support unstable joints. Use of a forearm crutch on the side of the body opposite the arthritis reduces forces on a hip or knee by as much as 50% [35]. No orthosis is available for hip problems, but braces can support unstable knees [35]. A wedged insole designed to compensate for valgus stress has been used in patients with mild osteoarthritis of the medial knee compartment, and in at least one observational study the insole reduced pain and improved function [36]. We identified no controlled trials involving the wedged insole or any other assistive devices.

Although convincing evidence from one observational study [37] showed that weight loss reduces the risk for developing symptomatic knee osteoarthritis in women, we found no controlled trial of weight loss as a treatment for osteoarthritis.

Although vibration therapy has been evaluated for several chronic pain syndromes [38], we identified no trials that specifically addressed osteoarthritis.

Discussion

Our literature search for trials of nonmedicinal, noninvasive therapies for hip or knee osteoarthritis yielded various potential therapies but a small number of trials. Of the 600 citations identified by our search of the literature of the past 28 years, only 31 (5%) articles addressed such therapies compared with 377 (63%) articles that evaluated nonsteroidal anti-inflammatory drugs. Trials that met our inclusion criteria for any single intervention were few, from 0 to 4.

Of the seven therapies for which data addressed efficacy in osteoarthritis of the knee or hip, exercise had the strongest evidence for a beneficial effect. In the three exercise trials with nonexercise controls [9, 10, 12], patients assigned to the exercise groups had greater pain reduction and more improved function than did the control groups. In the fourth study [13], in which all groups performed some exercise, the addition of aerobic activity to other exercises improved function but did not further reduce pain. Exercise programs evaluated varied greatly, from simple leg lifts done at home [9] to a sophisticated supervised program that included education, strengthening, and aerobics [10]; none precipitated disease flares. Both studies that included aerobic exercise [10, 13] indicated that patients with moderate to severe knee osteoarthritis can improve their aerobic capacity and exercise ability by participating in activities equivalent to 30 minutes of walking or swimming three times a week. No evidence showed that the addition of aerobic exercise improves the pain associated with osteoarthritis to a greater degree than that achieved by stretching and strengthening alone; however, only one study specifically addressed this question. Three of the exercise studies included knee osteoarthritis only. We found no data regarding the efficacy of exercise specifically for hip osteoarthritis. Although exercise studies showing a beneficial effect may be more likely to be published than negative studies (publication bias), these studies, along with the extensive exercise literature [39], strongly suggest that stretching and strengthening exercise alone or combined with aerobic walking or swimming reduces pain and improves function in patients with knee osteoarthritis. Aerobic exercise may offer some additional improvement in function over stretching and strengthening alone. More studies are needed to address the degree of education necessary to achieve compliance with treatment (handouts or intermittent visits to exercise specialists), the optimal type of exercises and their frequency, and the treatment of painful joint flares (apparently an unusual occurrence during exercise programs). In addition, more study is needed to evaluate the efficacy of exercise to treat osteoarthritis of the hip and other joints.

Diathermy (including ultrasound) before exercise is expensive and time-consuming. Available evidence suggests that this treatment provides no benefit in terms of pain reduction or improvement in function when added to an exercise program. The study by Falconer and colleagues [8] in particular provides strong support for lack of substantial benefit of ultrasound. Although other forms of diathermy may differ from ultrasound, no differences in pain or function between patients with knee osteoarthritis treated by ultrasound or by short-wave diathermy were shown [9, 40].

The data were insufficient to evaluate the efficacy of the other five nonmedicinal, noninvasive therapies, especially considering the possibility that positive results of trials may be more likely to be published than those not showing a beneficial effect. A single trial on topically applied capsaicin suggests that it may be useful to reduce pain associated with knee osteoarthritis. However, other trials are needed to confirm this result and should determine whether any change in function associated with such treatment also occurs. Other plant derivatives, such as menthol, that are commonly used by patients should also be evaluated.

Although we identified two trials of laser therapy, the second trial included too few patients with osteoarthritis, and thus evaluation of efficacy of this treatment is based on the trial of Stelian and colleagues [17]. This single, well-designed study suggests that laser treatment may be useful for reducing the pain and disability associated with knee osteoarthritis. However, the unusually strong intervention effect and weak placebo effect reported suggest the need for additional evaluation to confirm these findings.

We found only two trials of acupuncture: One reported no difference between acupuncture and sham treatments (needles placed at nontraditional sites), and the other reported benefit of acupuncture compared with no treatment. These results are consistent with a strong placebo effect of such treatment, an effect greater than placebo due to needle therapy not necessarily restricted to acupuncture sites, or an effect of acupuncture confined to the knee, which could have been obscured in the study by Gaw and colleagues [11]. We found no large single, well-designed studies of acupuncture; the existing studies give inconsistent results.

The three studies of transcutaneous electrical nerve stimulation therapy all reported superior pain control of active treatment compared with sham treatment. However, strong placebo effects occurred in all three studies. Only one study used an objective measure of lower extremity function, but it had inadequate numbers to evaluate because of the many participants who dropped out (34%).

One small study [18] on pulsed electromagnetic fields suggests that this therapy may be useful to reduce pain associated with knee osteoarthritis. However, more patients must be evaluated.

The degree of improvement reported in several of the trials of nonmedicinal, noninvasive therapies is similar to that associated with nonsteroidal anti-inflammatory drugs [6]. Why are there so few trials of such promising alternative therapies? Studies of safety and efficacy are required of drug manufacturers before Food and Drug Administration approval, but no such statutes exist for most alternative therapies. In contrast to drugs, there is often no clear financial sponsor of such studies of other interventions such as exercise. In addition, trials of alternative therapies are problematic because of the difficulty in masking the intervention or identifying appropriate placebo treatments and because of the difficulty in assuring patient compliance with regimens that may be more difficult to adhere to than medication. Patient compliance remains a formidable problem even when such therapies (for example, exercise) are effective. Many of these alternative therapies are expensive and time-consuming, especially those involving intensive patient contact, which may be responsible for part of the therapeutic effect.

However, the considerable risks associated with use of nonsteroidal anti-inflammatory drugs are being recognized now in older populations [41]. The risk for serious gastrointestinal side effects increases with nonsteroidal anti-inflammatory drug doses [4]. Thus, alternatives to these drugs or interventions that will allow use of lower doses would be welcomed by physicians who treat this common problem.