A total of 542 subjects were screened for eligibility for the study (Fig. 2) before the enrollment target of 200 eligible subjects was met. Twenty percent of screened subjects (108) declined to participate, while 234 did not meet the inclusion criteria. Of the 200 subjects enrolled in the study, 3 were lost to follow up: 2 in the Etoricoxib arm and 1 in the PSWT arm. These 3 subjects were considered dropouts, but included in the statistical analysis by assuming no changes in any measure (Fig. 2). All remaining (197) subjects were found to be compliant with the assigned treatment regime i.e. regularly taking prescribed dose of Etoricoxib or using the PSWT device continuously except for bathing.
Baseline demographics, COA grade, and baseline scores for NDI index and the two VAS measures are presented in Table 1. On average participants were experiencing neck pain for about 24 months and thus can be classified as chronic pain sufferers. These subjects reported a baseline NDI score of 24.5, indicating borderline severe disability and baseline pain scores during rest and activity exceeding 70 VAS mm, thus classifying them as severe pain sufferers.
Each of the measures was compared across the two arms. There were statistically significant differences across the two arms in age, NDI and the two VAS scores although in the latter three measures these differences were not clinically significant. In each case the PSWT arm reported, on average, less functionality and higher pain levels.
Table 1
Baseline demographic and clinical characteristics of the PSWT and NSAID arms. There were statistically significant differences across the two arms in age, NDI and VAS scores. However, the magnitude of these differences indicates no clinically meaningful differences in composition between the two treatment arms.
Baseline Demographic Data | All Subjects (N = 200) Mean (SD) | PSWT Arm (N = 100) Mean (SD) | NSAID Arm (N = 100) Mean (SD) | p value of PSWT Vs NSAID Means |
Age (yr) | 44.95 (10.3) | 46.40 (11.8) | 43.47 (8.3) | 0.04 |
Height (cm) | 164.9 (7.8) | 165.3 (7.9) | 164.47 (7.8) | 0.45 |
Weight (Kg) | 75.5 (14.8) | 76.2 (15.8) | 74.7 (13.8) | 0.47 |
BMI | 27.61 (4.0) | 27.72 (4.2) | 27.50 (3.9) | 0.7 |
Disease Duration (months) | 24.13 (24.1) | 24.4 (23.5) | 23.9 (24.8) | 0.88 |
Radiographic Imaging COA Grade | 2.69 (0.70) | 2.75 (0.73) | 2.63 (0.66) | 0.22 |
Gender (%) | Men | 28.5% | 35% | 22% | 0.04 |
Women | 71.5% | 65% | 78% | 0.04 |
VASrest (0-100) | 70.78 (9.9) | 72.57 (10.93) | 69 (8.66) | 0.01 |
VASactive (0-100 ) | 78.32 (9.4) | 79.95 (10.38) | 76.7 (8.15) | 0.01 |
Neck Disability Index (NDI) (0–50) | 24.57 (3.0) | 25.42 (3.77) | 23.66 (2.66) | 0.0002 |
Neck Disability Index (Functionality) and Visual Analog Scale (Pain)
After four weeks of treatment, subjects in both treatment arms reported lower mean scores for NDI, VASrest and VASactive) and these 4-week scores were statistically significant (p < .0001) from the relevant baseline measure (Table 2). The change in each component score is presented in supplemental Fig. 1.
Table 2
Subjects in both treatment arms reported a statistically significant reductions (p < 0.0001) in the three outcomes NDI, VASrest and VASactive when treatment scores were compared to their respective baseline score.
Outcome Measure | PSWT Arm Score Mean (SD) | NSAID Arm Score Mean (SD) |
Baseline | 4-weeks | Mean Difference | P value | Baseline | 4-weeks | Mean Difference | P value |
NDI | 25.42 (3.77) | 9.34 (4.46) | 16.08 (5.21) | p < .0001 | 23.66 (2.66) | 11.24 (4.68) | 12.42 (4.54) | p < .0001 |
VASrest | 72.57 (10.93) | 22.76 (12.78) | 49.81 (15.15) | p < .0001 | 69 (8.66) | 30.08 (14.60) | 38.92 (13.39) | p < .0001 |
VASactive | 79.95 (10.38) | 27.42 (14.81) | 52.53 (15.92) | p < .0001 | 76.7 (8.15) | 36.40 (16.40) | 40.3 (14.77) | p < .0001 |
To assess whether PSWT treatment was statistically superior to NSAID, 4 week treatment scores were compared between the two arms. In each of the three cases we find this difference is statistically significant at the p < 0001 level with subjects in the PSWT arm reported lower scores (Table 3, Supplemental Fig. 2). We also compared the mean difference in the outcome measures from the respective baseline and find these differences adjusted for baseline scores to be statistically significant at the p < 0001 level in favor of the PSWT arm. The standardized effect size was also calculated for the three outcome measures and the results can be seen in (Fig. 3) showing a large and consistent standardized effect size for the PSWT treatment.
Finally, to control for any differences in demographics as well as the baseline relevant outcome measures, we augmented our statistical analyses with ANCOVA analysis using ordinary least square regression and a dummy variable for the two arms. The dependent variables (DV) in these analyses were 4-week scores of the relevant outcome measure. The independent variables included a dummy variable associated with using the device (NSAID = 0; PSWT = 1), subject demographics (age, BMI, gender, duration of pain, OA grade) and the baseline score of relevant outcome measure.
The forecasted mean “adjusted” 4-week scores for the three measures and standard deviations of these adjusted mean are presented in the right hand column of Table 3, again indicating that PSWT treatment resulted in lower 4-week scores (i.e., was superior to NSAID treatment.) In addition this analysis found the coefficients for both BMI and baseline relevant measures (i.e., NDI/VAS) were statistically significant, indicating that subjects with higher BMI and higher baseline NDI/VAS scores reported less treatment efficacy at 4-weeks. No other variables were statistically significant in these analyses.
Table 3
The difference of mean differences between the two treatment arms: 1) as measured; and 2) after adjusting for demographics and baseline scores using ANCOVA. Statistical significance (p < 0.0001) was shown in both tests.
Outcome Measure | Measured 4- Week Score Mean (SD) | Comparison of Means (p value) | Adjusted (Regression) 4-Week Score Mean (SD) | Comparison of Means (p value) |
PSWT | NSAID | PSWT | NSAID |
NDI | 9.34 (4.46) | 11.24 (4.68) | p < .0001 | 9.00 (4.34) | 11.58 (4.34) | p < .0001 |
VASrest | 22.76 (12.78) | 30.08 (14.60) | p < .0001 | 21.98 (13.10) | 30.84 (13.10) | p < .0001 |
VASactive | 27.42 (14.81) | 36.40 (16.40) | p < .0001 | 26.44 (14.82) | 37.35 (14.82) | p < .0001 |
Rescue Medication Use & Treatment Satisfaction
Subjects in the NSAID arm used an average of 13.39 (10.80) dosages of rescue pain medication over the 4 weeks compared to only 6.73 (9.03) dosages for the PSWT arm or a difference of 6.73 (1.41) in favor of the PSWT (p < 0.0001). The distribution of rescue pain medication use was markedly different with 44% of the PSWT arm using no rescue medication compared to 13% of the NSAID arm. This difference is statistically significant (p < 0.0001). Subject satisfaction was found to be greater in the PSWT with an average rating of 76.39 (19.84) compared to an average rating of 59.55 (21.86) in the NSAID arm, a difference that is statistically significant (p = 0.0001)
Adverse Events
Adverse events (AEs) associated with NSAID or PSWT use were assessed during the study period and recorded at the end of the study (Table 4). In the NSAID arm, 2 subjects reported serious AEs of peripheral edema and hypertension, following which Etoricoxib treatment was ceased. There were 9 minor AEs in the NSAID arm, however these subjects chose to continue NSAID therapy after consulting with the PI. There were no AEs reported in the PSWT arm – the sole dropout in this arm did so at the beginning of the study, citing a preference for pharmacotherapy.
Table 4
Distribution and description of the Adverse Events in the two treatment arms.
Treatment Arm | Adverse Events/ Subjects | Event |
PSWT | 0/0 | 0 |
NSAID | 11/9 | • peripheral edema (n = 1) • gastric upset (n = 3) • hypertension (n = 5) • dysuria (n = 1) • increase in serum creatinine levels (1.5 mg/DL to 3.5 mg/DL) (n = 1) |
Dose-Response Characteristics (Post-hoc Analyses)
The three ANCOVA analyses reported above found treatment effectiveness (final 4-week score) was observed to be inversely proportional to subject BMI in all three cases. However, these analyses assumed the same loss of effectiveness for both treatment types as a function of BMI. The mechanism of action differs across the two arms. PSWT is a method of high-frequency magnetic stimulation, which utilizes a dipole magnetic field antenna. The decay of the magnetic field of the PSWT antenna (along the z-axis perpendicular to the antenna), Bz can be characterized as:
where R = radius of the antenna, Z = distance of the target away from the plane of the antenna and K is a constant. The subject PSWT device used in the present study has a radius of 6 cm.
The neck circumference is a useful predictor of BMI [38]. Moreover, the neck circumference can be used to calculate the diameter of the cross-section of the neck and thus estimate the depth of the PSWT target region (cervical spine) for various BMIs. For instance, a normal neck size (i.e. for a 20 BMI person) is about 38.1 cm, while a 30 BMI person’s neck size is about 43.2 cm. This implies that the cervical spine is located at a depth of 5.6 cm and 6.3 cm respectively from the skin for these two individuals. Inserting these values into Eq. 1 indicates that the field intensity at the cervical spine of an individual with a BMI of 20 would be about 21% greater than the field intensity for a person with a BMI of 30. As such, this leads us to postulate that the treatment effectiveness (analogous to dose responsiveness) of PSWT should be about 21% greater a 20 BMI person compared to a 30 BMI person.
To test this premise we conducted three separate univariate regression analyses for both the NSAID and PSWT arms, one for each of the outcome measures. The dependent variable (DV) in each regression is the reduction in relevant outcome measure (in this case baseline minus 4-week), and the independent variable is the subject’s BMI. In all six of these analyses we find the outcome measures were negatively correlated with BMI (Table 5). Using these coefficient estimates, it is possible to estimate the change in outcome measure score for various BMIs for each of the outcome measures. For instance, we calculate that a 20 BMI person using NSAID’s can expect to see a 43.69 points reduction in VASactive (52.75 − .453*20) = 43.69). However, if the person had a BMI of 30 the reduction would only be 39.16 points (52.75-.453*30). This implies the effectiveness (dose response ratio) of the NSAID is 12% (= 43.69/39.16) more effective when comparing a 20 BMI person to a 30 BMI person. Similarly, the figures for a person using PSWT would be a reduction of 57.85 VAS points if the person’s BMI was 20 and 50.98 if the person’s BMI was 30. This implies the device’s dose response ratio is 1.134 = 57.85/50.98 or 13.4% more effective for the 20 BMI person compared to the 30 BMI person (Table 5). The dose response ratio estimates for the NSAIDS vary from 1.07 to 1.21, mainly because the regression coefficient for BMI was not always significant. In contrast the estimate for dose response ratio for PSWT was very stable, ranging from 1.13 to 1.15 in line with our theoretical estimate of 23% derived earlier.
Table 5
Coefficients for the univariate regression of BMI for the various outcome measures.
Treatment Arm | Regression Coefficients |
Intercept | BMI |
Neck Disability Index (NDI) (0–50 points) |
NSAIDS | 19.21 | − .247 |
PSWT | 18.99 | --.105 |
| Visual Analog Scale (VAS)- Rest (0-100 mm) |
NSAIDS | 47.06 | − .296 |
PSWT | 70.10 | -732 |
| Visual Analog Scale (VAS)- Activity (0-100 mm) |
NSAIDS | 52.75 | − .453 |
PSWT | 71.58 | -687 |
Also investigated was the relationship between rescue medication use and the reduction in pain. We did this using regression analysis, with the DV being rescue medication use and the independent variables being standard demographics, baseline VASactive pain level and the change in VASactive pain level. The only statistically significant demographic variable was age, with older subjects using more rescue medication. However, both baseline pain and pain reduction were also significant, indicating that those with higher baseline pain used more medication and importantly those experiencing the largest pain reduction used less medication.