Specific-cytokine associations with outcomes in knee osteoarthritis subgroups: breaking down disease heterogeneity with phenotyping

Abstract

Background: Despite existing extensive literature, a comprehensive and clinically relevant classification system for osteoarthritis (OA) has yet to be established. In this study, we aimed to further characterize four knee OA (KOA) inflammatory phenotypes (KOIP) recently proposed by our group, by identifying the inflammatory factors associated with KOA severity and progression in a phenotype-specific manner.

Methods: We performed an analysis within each of the previously defined four KOIP groups, to assess the association between KOA severity and progression and a panel of 13 cytokines evaluated in the plasma and synovial fluid of our cohort's patients. The cohort included 168 symptomatic female KOA patients with persistent joint effusion.

Results: Overall, our analyses showed that associations with KOA outcomes were of higher magnitude within the KOIP groups than for the overall patient series (all p-values < 1.30e-16) and that several of the cytokines showed a KOIP-specific behaviour regarding their associations with KOA outcomes.

Conclusion: Our study adds further evidence supporting KOA as a multifaceted syndrome composed of multiple phenotypes with differing pathophysiological pathways, providing an explanation for inconsistencies between previous studies focussed on the role of cytokines in OA and the lack of translational results to date. Our findings also highlight the potential clinical benefits of accurately phenotyping KOA patients, including improved patient stratification, tailored therapies, and the discovery of novel treatments.

Keywords: Clinical severity; Cytokines; Inflammation; Knee osteoarthritis; Machine learning; Phenotype; Radiographic progression.

Fig. 1

Association of cytokines with knee osteoarthritis (KOA) severity and progression stratified by KOA inflammatory phenotypes (KOIP). The heatmap colours represent non-parametric correlation-like measurements to assess the association of the cytokines evaluated in our study with KOA outcomes, including clinical, radiographic and ultrasound severity at baseline, and radiographic progression at 2 years. Associations were assessed for the overall series and within each KOIP independently using Spearman correlations (continuous or ordinal outcomes) and Glass rank biserial correlations (binary outcomes). Baseline Kellgren-Lawrence staging and joint space narrowing were treated as ordinal in these analyses. Red indicates positive, blue represents negative, and colour intensity expresses more extreme values of the correlation coefficients. Colour intensities were saturated to 0.5 and − 0.5 for positive and negative correlation, respectively. IL-6, interleukin 6; IL-8, interleukin 8; TNF-alpha, tumour necrosis factor alpha; NGF, nerve growth factor; CRP, C-reactive protein; KOOS, Knee injury and Osteoarthritis Outcome Scores (reversed scores); KOA, knee osteoarthritis; KOIP, knee osteoarthritis inflammatory phenotype

Fig. 2

Association of synovial omentin with baseline pain measured by Knee injury and Osteoarthritis Outcome Score (KOOS, reversed score) within each Knee Osteoarthritis Inflammatory Phenotype (KOIP). The panels show the scatter plots for omentin and the KOOS scores in each KOIP group separately, the Spearman correlation coefficient, and its corresponding asymptotic 95% confidence interval (between brackets) and p-value. Omentin values are represented in a transformed scale according to Tukey’s ladder of powers, to symmetrize their distribution and make them more suitable for graphical representation (transformation parameter, g = 0.25); x-axis labels are shown in the original scale. Values from patients not belonging to the indicated KOIP group are represented in grey. Corr., correlation; pv, p-value; KOOS, Knee injury and Osteoarthritis Outcome Scores (reversed scores); KOIP, knee osteoarthritis inflammatory phenotype

Fig. 3

Association of synovial irisin with radiographic progression according to the Kellgren-Lawrence (KL) criteria within each knee osteoarthritis inflammatory phenotype (KOIP). The panels show the boxplots and stripcharts for irisin by patient groups in each KOIP group separately according to the patients’ progression status, as well as the group medians and their corresponding 95% confidence intervals. The legends display the fold changes (FC) between the groups, their 95% confidence intervals (between brackets), and the p-value for group comparisons derived from a Mann-Whitney test. A positive FC indicates higher average levels of irisin in progressors while negative FCs represent higher irisin levels in stable patients. Irisin values are represented in a transformed scale according to Tukey’s ladder of powers to symmetrize their distribution and make them more suitable for graphical representation (transformation parameter, g = 0); y-axis labels are shown in the original scale. FC, fold change; pv, p-value; KL, Kellgren-Lawrence; KOIP, knee osteoarthritis inflammatory phenotype

Fig. 4

Association of synovial interleukin 6 with joint effusion within each knee osteoarthritis inflammatory phenotype (KOIP). The panels show the scatter plots for interleukin 6 protein and joint effusion (mm) measured by ultrasound in each KOIP group separately, the Spearman correlation coefficient, and its corresponding asymptotic 95% confidence interval (between brackets) and p-value. Interleukin 6 protein values are represented in a transformed scale according to Tukey’s ladder of powers, to symmetrize their distribution and make them more suitable for graphical representation (transformation parameter, g = 0.25); x-axis labels are shown in the original scale. Values from patients not belonging to the indicated KOIP group are represented in grey. Corr., correlation; pv, p-value; KOOS, Knee injury and Osteoarthritis Outcome Scores (reversed scores); KOIP, knee osteoarthritis inflammatory phenotype