Participant characteristics
Of the 113 patients enrolled in the study, two dropped out because they were not willing to wait for the second EEG monitoring session after treatments. Considering the small dropout rate, recruitment was terminated with a sample of 111 patients, including 106 right-handed persons. The sex ratio was 36 men to 75 women. Among the 111 patients, 67 had significant pain relief after treatments (more than 50% decrease in NRS after treatments).
The average age of the 111 patients was 63±13 years, and their average BMI was 24.77±3.58 kg/m2. Table 1 shows the basic descriptive data and Table 2 shows the principal diagnoses of the patients.
Table 1 Basic descriptive data (n = 111)
Characteristics
|
Patients
|
Mean (SD)
|
Gender (M/F)
|
36/75
|
Age (y)
|
63(13)
|
Right-handed
|
106
|
BMI
|
24.77(3.58)
|
F, female; M, male; SD, standard deviation; BMI, body mass index.
Table 2 Principal diagnoses of the patients (n = 111)
Diagnoses
|
Number of patients
|
Osteoarthritis
|
38
|
Lumbar disc herniation
|
24
|
Postherpetic neuralgia
|
9
|
Other pain [1]
|
40
|
[1] Other pain included cervical spondylosis, tendinitis, back pain, cervicogenic headache and myofascitis.
Correlations between NRS and PI
Pearson’s correlations found a significant positive association (r = 0.800, p <0.05) between the 222 pairs of NRS (including NRSbefore and NRSafter) and PI (including PIbefore and PIafter) of the 111 patients before and after treatments (Table 3). Figure 1 shows the scatter plot of the data and the linear trend line for the association between NRS and PI (R2 = 0.64).
Table 3 Pearson’s correlations between NRS and PI
|
Pearson’s correlations
|
NRS and PI (n = 111)
|
0.800*
|
ΔNRS and ΔPI (n = 111)
|
0.721*
|
NRS-R and PI-R (n = 67)
|
0.840*
|
ΔNRS-R and ΔPI-R (n = 67)
|
0.664*
|
NRS-NR and PI-NR (n = 39)
|
0.782
|
ΔNRS-NR and ΔPI-NR (n = 39)
|
0.630
|
* p <0.05. PI, pain index; NRS, numerical rating scale; PI-R, PI of patients with significant relief after treatments (NRS scores decreased ≥ 50%); NRS-R, NRS scores of patients with significant relief after treatments (NRS scores decreased ≥ 50%); ΔNRS, the changes of NRS before and after treatments; ΔPI, the changes of PI before and after treatments; ΔNRS-R, the changes of NRS before and after treatments in patients with significant relief after treatments (NRS scores decreased ≥ 50%); ΔPI-R, the changes of PI before and after treatments in patients with significant relief after treatments (NRS scores decreased ≥ 50%); ΔNRS-NR, the changes of NRS in patients with no significant relief after treatments (NRS scores decreased < 50%); ΔPI-NR, the changes of PI in patients with no significant relief after treatments (NRS scores decreased < 50%).
A significant positive association was also found between the changes in NRS (ΔNRS) and the changes in PI (ΔPI) of the 111 patients (r = 0.721, p <0.05) (Table 3). Figure 2 shows the scatter plot and the positive linear trend for the association between ΔPI and ΔNRS (R2 = 0.52).
Association between PI and NRS in patients with significant pain relief after treatments
To determine whether PI was affected when pain changed significantly (as shown by NRS decreasing ≥50%), we analyzed the association between NRS and PI in patients with significant pain relief after treatments (NRS-R and PI-R, respectively). There was a significant positive association between NRS-R and PI-R (r = 0.840, p <0.05) (Table 3). The scatter plot of NRS-R and PI-R (Fig. 3) shows the linear trend for the relationship (R2 = 0.71).
A Pearson correlation of the changes in NRS and the changes in PI in patients with significant pain relief after treatments (ΔNRS-R and ΔPI-R, respectively) revealed a correlation coefficient of 0.664 (n = 67, p <0.05) (Table 3), indicating a strong positive association. The scatter plot of ΔNRS-R and ΔPI-R (Fig. 4) shows the linear trend for the relationship (R2 = 0.44).
Association between PI and NRS in patients with no significant pain relief after treatments
We found a significant positive association between NRS in patients with no significant pain relief (NRS decreasing <50%) after treatments (NRS-NR) and PI after treatments (PI-NR) (r = 0.782, p <0.05) (Table 3). The scatter plot of PI-NR and NRS-NR (Fig. 3) shows the linear trend for the relationship (R2 = 0.61).
The Pearson correlation of the changes of NRS-NR (ΔNRS-NR) and the changes of PI-NR (ΔPI-NR) before and after treatments revealed a correlation coefficient of 0.630 (n = 39, p <0.05), indicating a strong positive association. The scatter plot of ΔNRS-R and ΔPI-R (Fig. 4) shows the linear trend for the relationship (R2 = 0.40).
Furthermore, in order to determine whether the pain relief treatments affected PI in assessing pain, we compared the 95% confidence intervals of the R values of difference in NRS and PI in the significant pain relief group (0.6308 and 0.6972, respectively) and the no significant pain relief group (0.5985 and 0.6615, respectively). There was an overlap for these two intervals, indicating that there is no significant difference between the two correlation coefficients.
ROC analysis
The data from the 111 participants included 222 pairs of PI and NRS. NRS <4 were considered negative and NRS ≥4, positive. There were a total of 74 pairs in the negative group and 148 pairs in the positive group. We used ROC curves to analyze the sensitivity and specificity of PI to assess moderate-to-severe chronic pain (Fig. 5). The best fit with NRS categories was found with the following cut-off points based on the ROC curves: mild pain (NRS <4, PI <10.65), moderate pain (4 ≤ NRS < 7, 10.65 ≤ PI < 22.7), and severe pain (NRS >7, PI ≥22.7). From the ROC analysis, we found that the sensitivity of PI to NRS ≥4 was 80.4%, and the specificity was 86.5%, while the sensitivity of PI to NRS ≥7 was 100%, and the specificity was 93.9%. Table 4 shows the results of the ROC analysis.
Table 4 ROC analysis of PI and NRS
|
AUC (95% CI)
|
Cut-off value
|
Sensitivity
|
Specificity
|
NRS ≥ 4
|
0.912 (0.875, 0.948)
|
10.65
|
0.804
|
0.865
|
NRS ≥ 7
|
0.988 (0.973, 1)
|
22.7
|
1
|
0.939
|
AUC, area under the curve.
Associations of age, anxiety, and depression with PI
In this study, 63 patients completed the HADS. Multiple regression was conducted using PIbefore as the dependent variable (y) and using age and the anxiety (ANX) and depression (DEP) scales of the HADS as independent variables; age was x1, ANX was x2, and DEP was x3. The resulting model was y = 13.053 - 0.014x1 + 0.541x2+0.23x3, with an adjusted R2 of 0.138. Although the model (i.e., Model 1) was statistically significant (p <0.05), the low R2 indicated that the model was not a good fit, accounting for only a small proportion of the variance in PIbefore, and that PIbefore was significantly related to anxiety but not depression or age.
Given the results of Model 1, Model 2 tested the association of anxiety as the sole independent variable, with PI as y = 13.114 + 0.735x1. The adjusted R2 of Model 2 was 0.15, and it was not statistically significant. The results of the two models suggest that PI may be influenced more by anxiety than by depression, but the influence is small.