Overall, 564 potentially eligible studies were identified and 18 retrieved in full text.21 22 23 w1-w15
Two trials were not original21 22
and one made use of high intensity, continuous ultrasonography,23
leaving 15 eligible trials (fig 1).w1-w15
After adjustment for chance the agreement between reviewers (κ) on full text eligibility was 0.81 (95% confidence interval 0.68 to 0.94).
Fig 1Flow of trials through study. *Two sets of trials reported on common patient samples and were considered as single studies
Two trialsw11 w12 seemed to report on a shared group of 30 participants as the participants had an identical match for age. In addition, according to the methods sections these participants were recruited at the same institution over the same period. Attempts to contact the authors of these trials for clarification were unsuccessful. The data from both studies are considered as one trial for reporting purposes (fig 1).
Three trials by the same group of authors, published in three different journals in 2005, reported on the effect of low intensity pulsed ultrasonography on lateral malleolar fractures.w7-w9 Contact with the lead author confirmed that there were two distinct trials: one randomised trial of 22 patientsw7 and another of 30 patients.w8 w9 This left 13 unique trials for analysis (fig 1).
Most studies reported only surrogate end points; five explored end points of importance to patients (table 1). Eleven trials used imaging methods to assess bone healing. Six studies used plain films,w1 w2 w6 w10-w13 one used dual energy x ray absorptiometry scans,w14 two used both dual energy x ray absorptiometry scans and multidetector computed tomograms,w7-w9 one used high resolution microradiographs of fixed biopsies,w5 and onew3 used sagittal computed tomography to assess scaphoid healing.
Eligible trials were of limited quality (table 2). Attempts were made to contact the authors of seven trials to resolve uncertainties,w3 w6-w10 w13 w15 and clarification was successful in four.w7-w10 w15 Eleven trials used a parallel design with random allocation of sham and active ultrasound devices; two did not use a sham device as their control.w3 w14 One trialw3 randomly assigned patients with non-operatively managed scaphoid fractures to usual care or to low intensity pulsed ultrasonography in addition to usual care, whereas another studyw14 randomly assigned one limb of patients who had undergone bilateral tibial osteotomy to ultrasonography and the other limb to control. Although one trialw13 stated that patients and care providers were blinded, the treatment and control devices were visually different. No study explicitly declared an intention to treat analysis, but no trials reported patient crossover and all patients were analysed according to the group to which they were randomly allocated. No trial described any cointerventions to which participants were exposed. Seven trials reported loss to follow-up, ranging from 3% to 47% (table 2), which in all cases was dealt with by excluding lost participants from both the numerator and denominator for all outcome calculations.
Table 2 Methodological quality of eligible randomised controlled trials
Trial ultrasound devices
In 12 of the eligible trials the treatment provided to the control group was indistinguishable from that provided to the treatment group, the exception being the trial in which the active and sham devices were similar but easily visually distinguishable.w13 In 12 of the 13 eligible studiesw1-w9 w11-w15 the investigators made use of the Sonic Accelerated Fracture Healing System (Exogen, Piscataway, NJ). The trials that used this device required their treatment groups to receive daily 20 minute sessions with an ultrasound signal composed of a burst width of 200 μs (SD 10%) containing 1.5 MHz (SD 5%) sine waves, with a repetition rate of 1 kHz (SD 10%) and a spatial average temporal intensity of 30 mW/cm2 (SD 30%). The settings of the ultrasound unit could not be modified and a warning signal was sounded for active devices if coupling to the skin was not achieved. Duration of ultrasound use varied between trials: five studies instructed patients to apply low intensity pulsed ultrasonography until their fracture was healed,w3 w4 w6 w10 w14 seven used a set time that ranged from 13 hours to 90 days,w2 w5 w7-w9 w11-w13 w15 and in one the patients applied low intensity pulsed ultrasonography up to a maximum of 140 days until their fracture was healed (table 1).w1
Five trials reported on patient compliance with low intensity pulsed ultrasonography.w1 w2 w5 w7 w11 w12 Compliance was measured by an elapsed time recorder that provided only the total time used and not the temporal picture of use and by a daily log book maintained by participants. All found high agreement between the internal device timer and patients’ log books, and that use of the device between treatment and control groups was not significantly different. One of the trialsw5 provided additional details on use of the device, noting that although the patients applied low intensity pulsed ultrasonography on a daily basis, treatment with the active device was interrupted in 11% of applications owing to disconnected cables, improper contact between transducer and skin, or a low battery; however, patients successfully corrected the error and resumed treatment in all cases but one.
One studyw10 used an alternate ultrasound device, the Theramed 101-B ultrasound device supplied by the Instituto Nacional de Investigaciones en Metrología (Havana, Cuba). The signal intensity was 30 mW/cm2, and the device was described as low intensity pulsed ultrasound therapy. Patient’s applied this device for 20 minutes each day until radiographic healing, and active and sham units were blinded in the same manner as the Exogen device. None of the 13 eligible trials reported any adverse reactions or complications attributable to the device.
When time to radiographic healing—the most commonly reported end point among eligible trials—was pooled across all studies it showed a moderate effect in favour of low intensity pulsed ultrasonography. The pooled mean reduction in radiographic healing time was 33.6% (95% confidence interval 21.4% to 43.8%) but the associated heterogeneity was high (I2=76.9%; heterogeneity P<0.01; fig 2). Tests of interaction provided no evidence to support a different treatment effect across clinical presentations. The effect of low intensity pulsed ultrasonography was not significantly different between conservatively managed fresh fractures and operatively managed fresh fractures (P=0.48), between conservatively managed fresh fractures and operatively managed non-unions (P=0.61), or between operatively managed fresh fractures and operatively managed non-unions (P=0.39).
Fig 2Effect of low intensity pulsed ultrasonography on radiographic healing of fractures
Table 3 presents a detailed GRADE description for the effect of low intensity pulsed ultrasonography on return to function or acceleration of radiographic healing of non-operatively managed fresh fractures, non-operatively managed stress fractures, operatively managed non-union, and operatively treated fresh fractures. Trials addressing distraction osteogenesis are not shown as they did not report any functional outcomes or any common surrogate end point.
Table 3 GRADE evidence profile: randomised controlled trials of low intensity pulsed ultrasonography for more rapid return to function (often measured by surrogate of radiographic fracture healing)
Non-operatively managed fresh fractures
One study found no effect of low intensity pulsed ultrasonography on conservatively managed, isolated, clavicle shaft fractures.w15 Subjective fracture consolidation among patients treated with low intensity pulsed ultrasonography occurred in a mean 26.8 days compared with 27.1 days in the control (mean difference 0.3 days, 95% confidence interval −5.3 to 5.9), and no significant differences were found between groups in need for operative fixation, analgesic use, pain, adverse events, or resumption of sport, professional, or household activities. As patient assessed fracture healing, resumption of household activities, return to work, and resumption of sport measure the same underlying domain (functional recovery), a random effects model was used to pool data from these four end points to improve the precision of this outcome measure. The pooled standardised mean difference found that treatment with low intensity pulsed ultrasonography resulted in a non-significantly faster return to function by 1.4 days (95% confidence interval −0.6 to 3.4; I2=11.4%; heterogeneity P=0.34).
Low intensity pulsed ultrasonography significantly accelerated radiographic healing of fractures in all three trials that assessed this outcome.w1-w3
One trial found a 33.8% reduction in healing time, with distal radial fractures healing in 51 days compared with 77 days in the control group (mean difference 26 days, 95% confidence interval 6.4 to 38.6).w2
A second trial found a 30.3% reduction in healing time of scaphoid fractures; 43.2 days in the low intensity pulsed ultrasonography group compared with 62.0 days in the control group (mean difference 18.8 days, 7.6 to 30.0).w3
The authors did not specify the unit of their associated measures of variance, and for our analysis standard deviations were assumed. A third trial found a 46.3% reduction in healing time of tibial shaft fractures: 102 days in the low intensity pulsed ultrasonography group compared with 190 days in the control group (mean difference 88 days, 50.4 to 125.6).w1
This trial also found a significant improvement in surgeon assessed clinical healing (fracture stable and not painful to manual stress) of 86 days compared with 114 days (mean difference 28 days, 4.9 to 51.1), but not in time to partial weight bearing (45 days in the low intensity pulsed ultrasonography group v
49 days in the control group; mean difference 4 days, −11.0 to 19.0).15
The pooled results from the three trialsw1-w3 found a significant mean reduction in radiographic healing time of 36.9% (95% confidence interval 25.6% to 46.0%; I2=41.6%; heterogeneity P=0.18; fig 2). Calculating a Δ of 20% relevant to the control data for each trial (15, 12, and 38 days) and using the standard deviation associated with fracture healing in the three studies (31.6, 15.8, and 37.6 days) yields corresponding required sample sizes of 140, 56, and 32. The 158 patients available for this analysis therefore meet the optimal information size. Low quality evidence from three trials suggests a benefit of low intensity pulsed ultrasonography in non-operatively managed fresh fractures (tables 2 and 3).w1-w3
Non-operatively managed stress fractures
One studyw4 noted no improvement in return to full participation and duty among midshipmen sick listed because of tibial stress fractures. Patients treated with low intensity pulsed ultrasonography returned to active duty in a mean 55.8 days compared with 56.2 days for those receiving sham therapy (mean difference 0.4 days, 95% confidence interval −13.4 to 14.2). This trial provided moderate quality evidence of no effect of low intensity pulsed ultrasonography on return to function among non-operatively treated stress fractures (tables 2 and 3).
Operatively managed distraction osteogenesis
One studyw5 found no effect of low intensity pulsed ultrasonography in the stimulation of bone formation in the distraction gap created in severely resorbed mandibles, and suggested that future trials consider a longer consolidation period than 31 days. Another studyw6 found that low intensity pulsed ultrasonography accelerated radiographic healing in patients with tibial defects managed with distraction osteogenesis. The authors used a “healing index” as their outcome, which was defined as the duration of external fixation divided by the length of distraction gap. Patients using active low intensity pulsed ultrasonography had a healing index of 30 days/cm compared with 48 days/cm for those exposed to a sham device (mean difference 18.0 days/cm, 95% confidence interval 11.7 to 24.3). One studyw14 administered low intensity pulsed ultrasonography or sham treatment to patients undergoing opening wedge high tibial osteotomy to tackle varus deformity secondary to osteoarthritis. The authors noted that low intensity pulsed ultrasonography compared with sham treatment resulted in a significant increase in mean bone mineral density in the distraction callus (0.20 g/cm2 v 0.13 g/cm2; mean difference 0.07 g/cm2, 95% confidence interval 0.003 to 0.14) but not in bone mineral density distal to the distraction gap, nor in the mean consolidation period or duration of external fixator use. The authors did not specify the unit of their associated measures of variance, and for our analysis these were assumed to be standard deviations. Three trials provided very low quality evidence for accelerated functional improvement after distraction osteogenesis (table 2).w5 w6 w14
Operatively managed (bone graft) non-union
In one studyw10 the application of low intensity pulsed ultrasonography to patients with established scaphoid non-union and treated with vascularised pedicle bone graft compared with those exposed to a sham device accelerated healing by a mean difference of 38 days (95% confidence interval 26.3 to 49.7), which represents a 40.4% (95% confidence interval 30.8% to 48.7%) reduction in healing time (fig 2). Communication with the author established that the reported associated measures of variance were standard errors of the mean, and these were converted to standard deviations. To be considered healed, patients had to present with no tenderness at the scaphoid and show complete bridging of cortices on plain radiographs. Calculating a Δ of 20% relevant to the control data (94×0.2=19 days) and using the standard deviation associated with fracture healing (27 days) yielded a required sample size of 64. The 21 patients available for this analysis therefore did not meet the optimal information size. This trial provided low quality evidence for a benefit of low intensity pulsed ultrasonography in accelerating healing of established non-unions managed with bone graft (tables 2 and 3).
Operatively managed fresh fractures
Functional outcomes were inconsistent. One trial of patients with lateral malleolar fractures fixed using bioabsorbable screws found no differences in function at 18 months.w9 One trial of operatively managed tibial shaft fractures found no difference in time to full weight bearing (6.5 weeks for low intensity pulsed ultrasonography v 7.1 weeks for sham therapy; mean difference 0.6 weeks, −1.5 to 2.7).w11 w12 One trial of operatively managed tibial fractures reported that low intensity pulsed ultrasonography reduced average time to disappearance of site tenderness. Our reanalysis of the results found that this difference was not significant (6.1 weeks v 7.9 weeks; mean difference 1.8 weeks, −0.2 to 3.8, P=0.09). The authors did find that low intensity pulsed ultrasonography reduced time to full weight bearing (9.3 weeks v 15.5 weeks; mean difference 6.2 weeks, 4.4 to 8.0); there was no difference in time to partial weight bearing.w13
Two trialsw7-w9 of patients with lateral malleolar fractures fixed using bioabsorbable screws and treated with low intensity pulsed ultrasonography or a sham device found no significant differences in visualisation of the fracture line, external callus formation, percentage of bone healing, or bone mineral density. One studyw11 w12 found that low intensity pulsed ultrasonography had no effect on radiographic healing among patients with operatively managed (intramedullary nail) tibial shaft fractures. Active treatment resulted in a non-significant mean time to bridging of three cortices of 155 days compared with 125 days for sham treatment (mean difference 30 days, 95% confidence interval −16.5 to 76.5; fig 2). Leung et alw13 explored the effect of low intensity pulsed ultrasonography on operatively managed tibial fractures and found that it reduced time to removal of external fixator (9.9 weeks v 17.1 weeks; mean difference 7.2 weeks, 2.6 to 11.8) and time to first, second, and third callus formation; specifically, patients receiving active treatment showed formation of the third callus in an average of 11.5 weeks compared with 20 weeks for those receiving sham therapy (mean difference 8.5 weeks, 5.8 to 11.2), a 42.5% (95% confidence interval 31.7% to 51.6%) reduction in radiographic healing time (fig 2).
The pooled results from two trialsw11-w13 showed a non-significant mean reduction in radiographic healing time of 16.6% (−76.8% to 60.7%; I2=90.9%; heterogeneity P<0.01; fig 2). Four trials provided low quality evidence for acceleration of healing of operatively managed fresh fractures (tables 2 and 3).w7-w9 w11-w13