Journal of Rehabilitation Medicine 51-10 | Page 62

782
A. Jablochkova et al.
by principal components analysis (PCA), which is an unsu-
pervised statistical tool for data analysis and dimensionality
reduction, which helps evaluating the quality and homogeneity
of the dataset and identifies strong outliers using Hotelling’s T 2 ,
a multivariate equivalent of the 95% confidence interval (95%
CI). Secondly, OPLS-DA was used to confirm group compari-
son of the NGF, BDNF, and cytokine levels, and to investigate
the multivariate inter-correlations of NGF and BDNF with
cytokines and clinical variables. The statistical significance of
the regression was controlled using the p-value of the cross-va-
lidated analysis of variance (CV-ANOVA), which is diagnostic
tool for assessing model reliability. The p(corr) value was used
to investigate the importance of each variable in our MVDA
models. |p(corr)| ≥ 0.5 was considered significant. Quality of the
MVDA was controlled using R 2 and Q 2 . The R 2 value shows
how well the model explains the dataset. Cross-validated Q 2 is
a measure of the predictive power of the model; in other words,
it measures how well the model can predict a new dataset. In
a high-quality model, Q 2 and R 2 values are close to each other.
RESULTS
Baseline characteristics
Table I shows the background data of both groups. BMI
was significantly higher in patients with FM. There
Table I. Background data of women with fibromyalgia (FM) and
healthy female controls (HC)
Variables
Age, years
BMI, kg/m 2
Systolic BP, mmHg
Diastolic BP, mmHg
FM duration, years a
FM
n  = 75
Mean (SD)
50.8 (9.6)
27.4 (4.7)
125.1 (27.8)
78.9 (16.1)
HC
n  =  25
Mean (SD)
47.6 (12.8)
25.1 (4.8)
129.7 (14.7)
82.9 (7.8)
p-value
(p  < 0.05)
ns
0.041
ns
ns
Tender points, n a 9.6 (6.5)
15.8 (1.8) 0
0 < 0.001
< 0.001
VAS global pain (0–100) # 53.1(20.6) 0.2 (0.9) < 0.001
350.3 (131.7)
318.5 (102.5)
307.1 (131.8)
314.4 (123.9)
396.8 (167.9)
420.8 (139.9)
378.5 (176.1)
384.1 (189.2)
1.6 (1.9)
2.7 (2.7) < 0.001
< 0.001
< 0.001
< 0.001
< 0.001
< 0.001
< 0.001
< 0.001
< 0.001
< 0.001
PPT (kPa) a,b
Supraspinatus, right
Supraspinatus, left
Vastus lateralis, right
Vastus lateralis, left
Gluteus, right
Gluteus, left
Knee, right
Knee, left
HADS-depression (0–21)
HADS-anxiety (0–21)
PCS (0–52) a
189.7 (105.4)
197.8 (117.2)
153.8 (86.0)
150.6 (89.5)
190.7 (93.5)
194.6 (0.94.0)
182.2 (121.5)
183.2 (104.7)
7.5 (3.7)
8.9 (4.4)
SF36-PSC (0–100) b
SF36-MSC (0–100) b
FIQ (0–100)
MFI (0–20) b
General fatigue
Physical fatigue
Reduced activity
Reduced motivation
Mental fatigue
4.6 (8.6) < 0.001
29.5 (8.1) 48.2 (18.8) < 0.001
36.9 (11.5)
63.7 (14.7) 44.0 (18.1)
5.2 (6.3) 0.011
< 0.001
21.1 (11.3)
18.2
16.7
15.5
10.8
15.4
(2.2)
(2.7)
(3.4)
(3.8)
(3.7)
8.2
7.5
6.7
5.5
7.1
(3.2)
(3.2)
(2.9)
(2.4)
(2.9)
< 0.001
< 0.001
< 0.001
< 0.001
< 0.001
a
FM n  = 72. b HC n  = 24.
BMI: body mass index; BP: blood pressure; VAS: visual analogue scale; PPT:
pressure pain threshold; HADS: Hospital Anxiety and Depression Scales; PCS:
Pain Catastrophizing Scale; SF36-PSC: Short Form Health Survey 36 (SF36)
Physical Summary Component; SF36-MCS: Short Form Health Survey 36 (SF36)
Mental Summary Component; FIQ: Fibromyalgia Impact Questionnaire; MFI:
Multidimensional Fatigue Inventory. The significance of group differences is
presented in the right column as a p-value. NA: not applicable.
www.medicaljournals.se/jrm
were no significant differences in age and blood pres-
sures between FM and HC. FM and HC differed in the
psychometric variables: patients with FM reported sig-
nificantly worse health in all psychometric variables.
Baseline levels of nerve growth factor and brain-
derived neurotrophic factor
At baseline, there were significant differences in
plasma levels of NFG and BDNF between FM and HC
(Table II); FM had lower levels of circulating NGF than
HC (p < 0.001), whereas the levels of circulating BDNF
were higher in FM than in HC (p = 0.001).
Baseline levels of cytokines and chemokines
There were significant differences in IL-1β (p = 0.008)
and IL-8 (p = 0.023) concentrations between FM and
HC. FM had significantly higher IL-8, while IL-1β was
significantly lower (Table II).
Group differences at baseline in a multivariate
context
To evaluate the importance of the bivariate statistical
comparisons, including their relative importance, we
performed an OPLS-DA of group membership (FM or
HC) using NGF, BDNF, and the 13 cytokines and che-
mokines as regressors (i.e. x-variables reported in Table
II). BDNF (p(corr) = –0.81) and NGF (p(corr) = 0.51)
turned out to be the most important variables of this
model representing the difference between FM and
HC (R 2  = 0.38, Q 2  = 0.24, CV-ANOVA: p < 0.001). All
the cytokines had |p(corr)|< 0.5 and were thus not of
significant importance in the multivariate context for
differentiating between FM and HC.
Table II. Nerve growth factor (NGF), brain-derived neurotrophic
factor (BDNF), cytokine, and chemokine levels at baseline from
women with fibromyalgia (FM) and healthy female controls (HC)
Substance, pg/ml FM
n=75
Median (IQR) HC
n=25
Median (IQR) p-value
NGF
BDNF 0.28 (0.33)
0.69 (0.65)
1,553.30 (2323.80) 671.6 (2250.2) ≤ 0.001
0.001
IFN- ɣ
IL-1β
IL-2
IL-4
IL-6
IL-8
IL-10
IL-17A
TNF-α
IL-1ra
IP-10
MCP-1
Eotaxin
10.01 (12.53)
0.07 (1.62)
2.9 (4.01)
0.56 (8.63)
1.83 (2.53)
1.37 (1.96)
9.24 (10.2)
10.10 (11.82)
4.31 (2.41)
35.6 (121.22)
398.53 (252.26)
309.27 (126.43)
107.6 (49.10)
8.45 (9.09)
1.23 (1.81)
2.25 (3.17)
0.56 (5.97)
1.68 (1.55)
0.77 (1.51)
8.7 (4.78)
9.7 (8.77)
3.64 (1.46)
35.83 (74.01)
395.36 (228.40)
336.1 (138.54)
106.89 (49.90)
IQR: interquartile range. Significant values are shown in bold.
0.405
0.008
0.602
0.805
0.222
0.023
0.867
0.173
0.335
0.981
0.424
0.292
0.186