Two prospective Swedish observational cohorts formed the basis of this study. Clinical patient characteristics are outlined in Table 1. The TIRx (extra-early rheumatology follow-up) cohort represents a population with increased risk of developing RA. Inclusion criteria were positive ACPA test (anti-CCP2) in clinical routine care and musculoskeletal pain of any kind and duration (10). Exclusion criteria were previous inflammatory rheumatic disease or corticosteroid treatment (oral or intra-articular) within 6 weeks prior to screening. A total of 82 patients without arthritis upon clinical examination at baseline were included, thus representing an at-risk population. During a median of 6 years follow-up, 39 (48%) of the patients developed arthritis, defined by clinical examination by an experienced rheumatologist (10).

We also studied 412 early RA patients from the TIRA-2 (Timely Interventions in RA) cohort (11). Inclusion criteria were symptom duration ≥6 weeks but <12 months since the first joint swelling as judged by the patient. In addition, patients should fulfill four of seven of the 1987 revised American College of Rheumatology criteria for RA (12) (n = 387 [94%]) or experience morning stiffness for ≥60 min and symmetrical arthritis and small joint arthritis (fingers, wrists, or toes) (n = 25 [6%]). Radiographic damage was graded according to the Larsen method (13).

No participating individual had received disease-modifying anti-rheumatic drugs (DMARDs) prior to the initial blood sampling (baseline visit).

As controls, we recruited 100 blood donors who were age-matched for TIRx [mean age 55 (range 18–72) years, 50% women] from the Department of Transfusion Medicine at Linköping University Hospital.

Ethical permission has been granted; EPN-Linköping Dnr M168-05 (TIRA2), M220-09 (TIRx), and 2015/236-32 (controls).

Serum samples, stored at −70°C until analysis, were analyzed in duplicates for serotonin by enzyme-linked immunosorbent assay (ELISA) according to the manufacturer’s instructions (ImmuSmol, Bordeaux, France).

Erythrocyte sedimentation rate (ESR), C-reactive protein (CRP) and platelet count were analyzed according to clinical routine practice at the respective participating rheumatology unit.

Mann–Whitney U test and Kruskal Wallis were used to test differences in levels between groups and Pearson’s chi-square was used for dichotomous variables. For correlation analyses, Spearman’s correlation was applied. Cox regression was used to test serotonin levels versus (vs.) progression to arthritis. Statistical analyses were performed with SPSS statistics version 26 (IBM, Armonk, NY, USA) or GraphPad Prism version 9 (GraphPad Software, La Jolla, CA, USA). Two-sided p-values < 0.05 were considered significant.

Baseline serum serotonin levels were significantly higher in at-risk individuals (median 111 ng/ml) compared to healthy controls (median 65 ng/ml) p < 0.0001 (Figure 1A). The levels were comparable between patients progressing (median 127 ng/ml) vs. not progressing (median 102 ng/ml) to arthritis (Figure 1B). Patients with early RA also had significantly higher levels (median 135 ng/ml) compared to healthy controls (median 65 ng/ml, p < 0.0001; Figure 1B). No significant differences were found between at-risk individuals and patients with early RA. Neither were there any differences when comparing patients with early RA vs. at risk-individuals progressing to arthritis, or vs. at-risk individuals not progressing to arthritis (Figure 1B).

Longitudinal analyses in at-risk individuals revealed relatively stable serotonin levels over time (Figure 2A). At-risk patients progressing to arthritis had numerically higher median levels compared to those not progressing at all time points, but without statistical significance (baseline: median 127 vs. 102 ng/ml; 3 months; 147 vs. 115 ng/ml, time-point for arthritis development vs. 12 months; 112 vs. 97 ng/ml). Among progressors, serotonin levels at baseline (median 127 ng/ml) were similar to those at the time for arthritis development (median 112 ng/ml) (Figure 2B). Serotonin levels were not prognostic for arthritis development, as analyzed by Cox regression (p = 0.062).

In at-risk individuals, no statistically significant difference in baseline serotonin levels were found between patients with normal (n = 76, median 108 ng/ml) vs. raised CRP (n = 6, median 136 ng/ml) p = 0.36. Patients with raised ESR had lower serotonin (n = 21, median 73 ng/ml) compared to patients with normal ESR (n = 60, median 122 ng/ml), but not significant p = 0.055. Our serotonin analysis detects total serum content, including a potential proportion from platelets. Therefore, we investigated the association between serum serotonin and platelet count (rho = 0.15, p = 0.18). A total of 11 patients have elevated platelet count (i.e., above the reference range), but no significant difference regarding serotonin was seen when comparing patients with high (median 142 ng/ml) and normal (median 109 ng/ml) platelet count, p = 0.08.

No correlations were found between serotonin levels and CRP over time among at-risk individuals; baseline CRP (rho = 0.10, p = 0.38), after 3 months (rho = −0.15, p = 0.20) or after 12 months (rho = 0.14, p = 0.38). Analyzing ESR and serotonin over time revealed a significant correlation after 3 months (rho = −0.23, p = 0.04). Otherwise, no correlations were found; baseline ESR (rho = −0.13, p = 0.25) and after 12 months (rho = −0.12, p = 0.48).

In early RA, no correlations were found between baseline serotonin and disease activity (DAS 28) over time; Baseline DAS 28 (rho = −0.016, p = 0.75), after 3 months (rho = 0.008, p = 0.89) or after 36 months (rho = −0.002, p = 0.97). Serotonin did not correlate with baseline ESR (rho = 0.009, p = 0.86) or CRP (rho = −0.010, p = 0.84). There was no correlation between serotonin and baseline erosion defined by the Larsen score (rho = −0.018, p = 0.72) or after 36 months (rho = 0.077, p = 0.23). No differences in serotonin were seen in early RA patients positive vs. negative for ACPA (median 140 vs. 128 ng/ml, p = 0.33).

At risk-individuals prescribed non-steroidal anti-inflammatory drugs (NSAIDs) at their baseline visit (n = 31) had slightly higher baseline serotonin levels (median 127 ng/ml) compared to those who were not prescribed NSAIDs (n = 51, median 97 ng/ml), but this was not statistically significant (p = 0.059). However, serotonin levels at 3 months among individuals with NSAID prescribed at their baseline visit were higher compared to those without prescription (152 ng/ml vs. 105 ng/ml, p = 0.012).

In the early RA cohort (143 prescribed vs. 264 not prescribed at baseline), this difference was borderline significant (median 145 ng/ml vs. 130 ng/ml, p = 0.05). When combining both cohorts, patients who were not prescribed NSAIDs still had significantly higher baseline serotonin levels compared to controls (median 127 ng/ml vs. 65 ng/ml, p < 0.0001).

No correlations were found between baseline serotonin levels and visual analog pain scales (VAS Pain) at baseline (rho = 0.039, p = 0.44), after 3 months (rho = −0.037, p = 0.49) or after 36 months (rho = 0.071, p = 0.20) in early RA. In at-risk individuals, we only had access to patient global assessment VAS, which did not associate with baseline serotonin levels (rho = 0.073, p = 0.52).