47 patients with non-septic knee arthritis were treated by local ice (30 min) or cold CO2 (2 min) twice at an 8 hour-interval for 1 day.

The synovial fluid was collected just before the first cold application then 24 hours later.

Cytokine, VEGF, NF-kB, PG-E2 levels were assessed in the synovial fluid before/after treatment.

Contralateral arthritic knees were used as paired controls when possible.

Patient inclusion Patients hospitalized in the rheumatology department in the Besançon university hospital in France and suffering from non-septic knee arthritis (rheumatoid arthritis according to the ACR-EULAR criteria, spondyloarthritis according to the ASAS criteria, gout or calcium pyrophosphate deposition disease (CPDD) - diagnosed by microscopic microcrystal assessment in synovial fluid) were included consecutively after signed informed consent. The protocol was declared and approved by the local ethic committee - clinicaltrials.gov:NCT02573298, Comité de Protection des Personnes - Est II: 12-664) and all research was performed in accordance with relevant guidelines/regulations. Patients suffering from septic arthritides and knee osteoarthritis were excluded. The patients had received no biologic treatment nor conventional DMARD for the 6 months preceding inclusion. Corticosteroids, colchicine and NSAIDs were stopped for at least 24 hours prior to inclusion.

Study design In the first phase of the study, the included patients were then randomized (1:1) to receive either local ice (Thermogel®, Artsana, Grandate, Italy - 30 minutes application - N=16) or hyperbaric cold CO2 at -78°C (Cryo+®, Cryonic, Salins-les-Bains, France - 2 minutes-N=16). Each patient received two applications of the randomly chosen technique at an 8 hour interval (9 a.m and 5 p.m). The skin temperature was monitored on the treated knee using MLT409/A Skin Temperature Probe® and ML309 Thermistor Pod® (ADInstruments,Dunedin, NZ). Thirty of these 32 patients were also included in another part of the study,aiming at evaluating the variations of the synovial power Doppler semi-quantitative score before/after 2 cold applications (ice versus cold CO2) 2. Just before the first cold application, at 9 a.m., and 24 hours later (day=1 at 9 a.m), an arthrocentesis was performed. Standard analyses were performed on the synovial fluid (bacteriology, cytology and microcrystal microscopic assessment). Furthermore, a part of the synovial fluid was centrifuged then frozen at -80°C. For the second arthrocentesis, after the synovial fluid was gathered for the same analyses, an intra-joint corticosteroid injection (Triamcinolone, HEXATRIONE®, Ethypharm, Saint-Cloud, France) was performed before removing the needle. These synovial fluid samples were used to perform the present part of the study, which was overall powered to evaluate the IL-6 level variations in the synovial fluid before/after 2 cold applications. After all the patients were included, synovial fluid IL-6, IL-17A, IL-1β, TNF-α, VEGF (Multiplex flow cytometry, CBA® BD Bioscience, Franklin Lakes, New Jersey, USA) and PG-E2 (ELISA, KGE004B®, Bio-Techne, Minneapolis, Minnesota, USA), NF-KB-P65/NFkB- p65-P (ELISA,85-86083-11®, Thermofisher, Waltham, Massachusetts, USA) levels were measured.

In the second phase of the study, we only included patients suffering from arthritides of both knees and treated them with local ice only, according to the protocol described above (N=15 +1 patient with knee bi-arthritis previously included in the first phase of the study in the ice treated group). The same protocol was applied to contralateral non-treated knees except cryotherapy treatment. Therefore, the synovial fluid was gathered and analyzed at the same evaluation times compared to treated knees, so these contralateral arthritic knees were used as paired controls for cytokine and enzyme assays (N=16).

Statistical analyses. The sample size was calculated in order to detect a significant variation in IL-6 synovialprotein levels before/after 2 cold applications. 15.78 (N=16) patients per group were necessary to detect a difference of 2325 pg/mL in IL-6 protein level with a power of 95% and a p-value of 0.05, according to published results of IL-6 assays in knee synovial fluid. Therefore, 2 groups of 16 patients were included in the first randomized phase of the study(ice versus cold CO2), then 16 patients with knee bi-arthritis were required for the second phase of the study (ice-treated versus contralateral knee, N=15+1 patient already enrolled in the first phase). For these reasons, a total of 47 patients were included. Paired Wilcoxon Mann-Whitney tests were performed in order to compare the mean cytokine and enzyme levels before/after treatment. Subgroup analyses were also planned (according to the treatment modalities (ice or cold CO2) and to the type of rheumatic disease (microcrystalinduced arthritides - pooled gout and CPDD patients - versus non-microcrystal-induced diseases - pooled RA and SpA patients). Furthermore, an interclass effect-size (weighted mean differences with 95% CI) for cytokine levels (before/after treatment) was calculated between ice-treated knees and the corresponding contralateral non-treated knees using R® software (rmeta® and meta® packages). Correlation tests were also performed using Pearson's coefficients in order to assess the parameters associated with cytokine level variations (before/after treatment). The statistical analyses were performed using R® and Graphpad® softwares.

Additional informations Funding: This work was supported by GIRCI Est II ("Young scientist" grant - 2014 - 21324 euros).

Competing interests: None.

Inclusion Criteria:

• non-septic knee arthritis

Exclusion Criteria:

• biologic treatment within the last 6 months, septic arthritis

the cytokine level assessment was blinded regarding the treatment arm