Forty-nine patients with FM formed the study population. Patients were randomized either to a moderate- to high-intensity Nordic Walking (NW) program (n = 26) or a supervised low-intensity walking (LIW) program (n = 23). The supervised exercise programs were conducted twice a week for 40 to 45 minutes for 15 weeks. They participated in a blood test at baseline, after 15 weeks of exercise, and at 30 weeks of follow-up. The effects of Nordic walking on body function were reported previously 4 , showing that NW resulted in better improvement in the 6-minute walk test (6MWT) and aerobic capacity, when compared with patients in LIW. Pain did not significantly change in any of the exercise groups. For patient characteristics, see Table 1.

Criteria for inclusion: Women with FM, aged 20 to 60 years, and with interest in exercising outdoors twice a week for 15 weeks, to participate at blood test at baseline and after the exercise period. To ensure that the patients would manage the planned aerobic exercise, a bicycle test at 50 watts was included in the inclusion criteria. All included patients managed to perform the test. They were also invited to participate in an examination of cerebrospinal fluid, which, however, was not a criterion of inclusion. FM was defined by the ACR 1990 criteria 1 : a history of long-lasting generalized pain and pain in at least 11 of 18 tender points examined by manual palpation.

Criteria for exclusion: Patients not speaking or reading Swedish, other severe somatic or psychiatric disease, ongoing or planned physical therapy, including exercise, and inability to accept times for planned exercise sessions.

The pain threshold was examined by using an algometer (Somedic Production AB, Sollentuna, Sweden), measured in kilopascals (kPa) 37 . The pain threshold was measured in two tender-point locations in the upper and lower extremities, respectively. The mean value was applied, and a higher value indicates better health.

Pain was rated on a visual analog scale (0 to 100) of the Fibromyalgia Impact Questionnaire (FIQ) 38 and reflects subjective estimation of overall pain intensity during the last week. A higher score indicates more-severe pain.

The 6-minute walk test (6MWT) reflects physical restriction related to the pain disorder. The patient was instructed to walk as quickly as possible but not to run. The distance covered was measured in meters 39 .

Serum was collected at rest (n = 49) at baseline and after 15 weeks in the exercise program and at 30 weeks of follow-up (n = 40). Serum samples were acquired by venipuncture of the cubital vein.

Twenty-six patients agreed to participate in an examination of cerebrospinal fluid (CSF) at baseline, and 15 patients also participated after 15 weeks of exercise. CSF was collected through lumbar puncture through the L3/L4 interspace. Collected blood and cerebrospinal fluid samples were centrifuged at 800 g for 3 minutes, aliquoted, and stored frozen at -70°C until use.

Biologic markers were analyzed with enzyme-linked immunosorbent assay (ELISA) by using commercially available kits. Assays specific for IL-8 (DY208, detection limit 31 pg/ml), MMP-3 (DY513, 1.4 pg/ml), serum free bioactive IGF-1 (DY291, 4 pg/ml), and IGFBP3 (DY675, 0.125 ng/ml) were purchased from R and D Systems (Minneapolis, MN, USA). An assay specific for human IL-6 (M1916, 0.6 pg/ml) was purchased from Sanquin Reagents (Amsterdam, The Netherlands). Assays specific for SP (FEK-061-05, 1 pg/ml) and NPY (FEK-049-03, 1 pg/ml) were purchased from Phoenix Pharmaceuticals (Burlingame, CA, USA). The human NGF-specific assay was purchased from Promega (Madison, WI, USA; 4 pg/ml). All assays were run according to recommendations of the manufacturers. Ordinary colorimetric ELISA was read with a Spectramax 340 from Molecular Devices (Sunnyvale, CA, USA), and fluorescent ELISA assays were read with a Mithras LB940 from Berthold Technologies (Bad Wildbad, Germany).

The study was approved by the ethics committee of Sahlgrenska University Hospital (220-09). Written and verbal information was given to all patients, and written consent was obtained from all patients.

ClinicalTrials.gov NCT00643006.

Descriptive data are presented as median, interquartile range, the number, and percentage. The Mann-Whitney U test and Kruskal-Wallis test were used to analyze the differences in continuous variables between the groups. The mean of baseline and posttest values was calculated for each individual, and these means were used in group comparisons. The Wilcoxon signed-rank test was used for comparisons of continuous variables within groups. Relations between the variables were examined with the Spearman correlation coefficient. Partial correlational analyses were performed to adjust for baseline pain and age. To control for possible Type I errors, the upper limit of number of false significances was calculated by the following formula: (number of tests - number of significant tests on the level of α) × α/(1-α). All significant tests were two-tailed, and values of p < 0.05 were considered significant. All statistical evaluation of data was done by using the statistic program SPSS Statistics v. 19.

Baseline values for 6MWT, pain, and pain threshold are presented in Table 2. The 6MWT was found significantly to improve (P = 0.033) in the NW group (median increase, 36.5 m) when compared with the LIW (median increase, 8.0 m). Changes in pain levels and pain threshold did not differ significantly between groups (Table 2).

When calculated for the total group after 15 weeks of exercise, 6MWT was significantly improved (median, 25.1; P < 0.001), Table 3. Decrease of the pain threshold was significant (median, -9.6; P = 0.040), whereas the decrease of pain was not (median, -5.0; P = 0.052), Table 3.

A significant association was found between 6MWT and the pain threshold at baseline (r ^s = 0.294; P = 0.040). Change in 6MWT (Δ6MWT) was correlated with changes in pain thresholds (ΔPT) after 15 weeks (r ^s = 0.339; P = 0.020) and after 30 weeks (r ^s = 0.407; P = 0.006).

No significant between-group differences were found for changes in IGF-1 (P = 0.795) or the IGFBP-3 (P = 0.906) occurring during the 15-week low-intensity and high-intensity aerobic exercise, respectively.

In the NW group, IGF-1 did not significantly change during the 15-week exercise (P = 0.300), but it decreased at 30-week follow-up (P = 0.019). IGFBP-3 did not significantly change during the 15-weeks (p = 0.977) or during the 30 weeks (P = 0.426).

In the LIW group, IGF-1 did not significantly change during the 15 weeks (median, -0.340; P = 0.148), but it tended to decrease during the 30 weeks (P = 0.109). IGFBP-3 did not signficantly change during the 15 weeks (P = 0.881) or 30 weeks (P = 0.796).

When calculated for the total group, levels of either free IGF-1 or IGFB3 did not significantly change during the exercise period (Table 3).

Basal IGF-1 did not correlate with baseline pain (r^s = -0.026; P = 0.861). To evaluate the relation of basal IGF to the change in pain, we did a partial correlation between baseline IGF and final pain and adjusted for baseline pain. We found a tendency to positive correlation (r = 0.278; P = 0.059). However, this positive tendency disappeared when age was accounted for (r = 0.245; P = 0.101).

Change in IGF-1 levels (ΔIGF) was associated with ΔPT at 15 weeks (r = 0.276; P = 0.058), and the correlation reached a significant level at 30-week follow-up (r ^s = 0.449; P = 0.005). No significant correlations were found between ΔIGF-1 and change in rating of pain (Δpain) when analyzed for the total group.

A smaller group within the study population agreed to participate in the examination of cerebrospinal fluid at the start of the exercise period (n = 26). Their median age was 51 (46 to 55) years. At baseline, the number of tender points was 14.5 (12 to 16), pain threshold was 207 (161 to 258) kPa, pain level was 70 (44 to 83), and IGF-1 was 4.5 (2.2 to 9.3) ng/ml, implying that they did not significantly differ from the total sample.

Fifteen individuals were able to participate in a second examination of cerebrospinal fluid after 15 weeks. Their median age was 51 (34 to 59) years. At baseline, tender points for these 15 individuals was 15.0 (12 to 18), the pain threshold was 207 (177 to 255) kPa, the pain level was 70 (51 to 82), and IGF-1 was 4.2 (2.2 to 9.2) ng/ml, implying that they did not significantly differ from the total sample.

Changes in serum IGF-1 levels correlated positively with ΔCSF NPY (r ^s = 0.802; P = 0.001). The correlation between IGF-1 and ΔCSF SP tended to be significant (r ^s = 0.495; P = 0.072), Table 4.

Changes in pain threshold (ΔPT) were associated with ΔCSF SP (r ^s = 0.600; P = 0.023), and inversely with ΔCSF MMP-3 (r ^s = -0.569; P = 0.034), Table 4.

Baseline CSF IL-6 was positively correlated with CSF SP (r ^s = 0.438; P = 0.032) and CSF MMP3 (r ^s = 0.409; P = 0.047). Furthermore, baseline CSF IL-6 was inversely correlated with CSF NGF (r ^s = -0.410; P = 0.042; n = 25) and CSF NPY (r ^s = -0.399; P = 0.053; n = 24).

Analyses of within-group analyses and between-group analyses comprised 40 analyses (Tables 2 and 3), and the upper level of of number of false signifcances was 1.58, which means that two significances might be false. Analyses of correlations between IGF-1, pain, and CSF markers (Table 4) comprised a total of 18 analyses, and the upper level of the number of false significances was 0.79, which means that one of the significances might be false.