Table 4 Results of studies included in review

 Gebhardt et al., 2006 [23]

NSLBP vs. controls

VAS (last 24 h); Functional back capacity score

No significant differences in geometric mean hsCRP levels based on log-transformation between NSLBP and control at baseline (MD = 0.1 95%CI −0.6 to 0.7)).

No difference mean change from baseline to 6 months in the NSLBP group (MD = 0.1 95% CI − 0.6 to 0.7).

 Sturmer et al., 2005 [28]

NSLBP vs. controls

VAS (last 24 h)

There was no difference in hsCRP concentration between pain (> 4.5) compared to low values of pain (≤2.3) (OR = 0.87 (95% CI 0.25 to 3.0)) after adjusting for BMI, age, smoking, alcohol consumption, diabetes and analgesic drugs.

 Klyne et al., 2017 [25]

NSLBP vs. controls

VAS

CRP was higher in NSLBP participants than controls (p = 0.003).

Between the three groups, CRP levels were higher in those with high-pain (VAS ≥4) than low-pain (VAS < 4) groups (p = 0.005) and controls (post-hoc: p = 0.005).

Linear and quantile analysis revealed significant positive associations between CRP pain intensity (β = 0.17, 95%CI 0.03 to 0.31)

 Klyne et al., 2018 [26]

Recovered, partially recovered, unrecovered

VAS

RMDQ

CRP levels between the unrecovered group and recovered group (β = 2.47 (95% CI 1.09 to 3.85) and between the partially recovered and recovered group (β = 1.80 (95% CI 0.39 to 3.21) were significant.

 Heffner et al., 2011 [24]

NSLBP vs. controls

PSQI

MPQ-SF

There was no difference in IL-6 levels between NSLBP and control (MD = − 0.1 95%CI − 0.6 to 0.4)

IL-6 levels were also associated with higher MPQ-SF affective pain ratings (r = 0.46, p = 0.02).

Regression analysis showed IL-6 levels were not significantly related to pain (β =1.06; p = 0.14).

 Klyne et al., 2017 [25]

NSLBP vs. controls

VAS

There was no significant difference between NSLBP and controls for IL-6 (p = 0.141).

Between the three groups IL-6 was higher in high-pain (VAS ≥ 4) than the low-pain group (VAS < 4) (p = 0.034), but not the control group (p = 0.114).

 Klyne et al., 2018 [26]

Recovered, partially recovered, unrecovered.

VAS

RMDQ

 Luchting et al., 2016 [27]

NSLBP vs. controls

Increased serum levels of IL-1β in patients with neuropathic pain (p < 0.05) was found but not in NSLBP (p > 0.05).

 Klyne et al., 2017 [25]

NSLBP vs. controls

VAS

 Klyne et al., 2018 [26]

 Wang et al., 2010 [29]

NSLBP vs. controls

VAS (last 24 h and past week)

RMDQ

Median TNF-α serum levels of patients with chronic NSLBP (2.51 pg/mL) and NSLBP with depression (2.58 pg/mL) were significantly higher than age matched controls (0.1 pg/mL; p = 0.004 for NSLBP; p = 0.002 for NSLBP + depression).

No significant associations were found between TNF-α levels and pain intensity.

 Wang et al., 2008 [30]

NSLBP vs. controls

VAS

RMDQ

Significant difference between groups in percentage of subjects with elevated TNF-α (> 2 pg/mL) at all four time points (baseline OR = 9.5; 95%CI 5.0 to 18.2), (180 days OR = 5.7; 95%CI 3.0 to 11.0).

No significant association was found between levels of TNF-α with pain and disability scores.

 Klyne et al., 2017 [25]

NSLBP vs. controls

VAS

There was no significant difference between NSLBP and controls for TNF-α (p = 0.174).

There was no difference in TNF-α between the three groups.

Linear regression revealed significant an association between TNF-α with pain rumification (β = − 0.20, 95%CI − 0.37 to − 0.02).

 Klyne et al., 2018 [26]

Recovered, partially recovered, VAS

RMDQ–

TNF-α was lower in the recovered group at both time-points than the other groups. TNF-α was different between the unrecovered vs. recovered groups (β = −  0.68;95% CI − 1.08 to − 0.27) and partially recovery versus recovered (β = −  0.42; 95% CI−  0.72 to − 0.12).

 Luchting et al., 2016 [27]

NSLBP vs. controls

P2RX7 mRNA expression increased in patients with neuropathic pain to controls (MD = -0.6 95%CI − 0.9 to − 0.3), but not in patients NSLBP compared to controls (MD = -0.1 95%CI − 0.4 to 0.2).