Body mass index and extent of MRI-detected inflammation: opposite effects in rheumatoid arthritis versus other arthritides and asymptomatic persons

Three groups were studied. Firstly, RA patients who were consecutively included in the Leiden Early Arthritis Clinic (EAC) cohort between August 2010 and October 2014. Secondly, early arthritis patients with other inflammatory diagnoses who were included in the same cohort in the same time period. Thirdly, asymptomatic volunteers who were recruited from the general population.

The EAC cohort is an inception cohort of early arthritis patients presenting with clinically detected arthritis of ≥ 1 joint and symptom duration < 2 years [8]. At baseline, questionnaires were filled, physical examination was performed (including weight and height), blood samples were obtained, and MRI was performed. RA was defined as fulfilling the 1987 ACR criteria during the first year of follow-up.

The asymptomatic volunteers were recruited between November 2013 and December 2014 [9]. Volunteers were recruited via advertisements in local newspapers and websites. The volunteers had no history of RA or other inflammatory rheumatic diseases, no joint symptoms during the last month, and no clinically detectable arthritis at physical examination. Participants received a voucher of €20 to compensate for their time and travel costs and did not receive a report of the MRI results. Therefore, volunteers had no/limited benefit from participating.

On an ONI-MSK-extreme 1.5 T extremity MRI machine (General Electric, WI, USA), imaging was performed of the unilateral metacarpophalangeal (MCP) 2–5 joints, wrist joints, and metatarsophalangeal (MTP) 1–5 joints. In patients the most painful side was scanned or, in case of equally severe symptoms on both sides, the dominant side. In asymptomatic volunteers, the dominant side was scanned. According to the protocol, the MRI scan was performed before the 2-week visit in which patients receive their diagnosis, and the median time between the first visit and the MRI scan was 8 days. Furthermore, patients were asked to stop NSAIDS 24 hours prior to the MRI scan. The following sequences were acquired: T1 fast-spin echo (T1), T2-weighted fat saturated (T2), and, after intravenous contrast administration (gadoteric acid, 0.1 mmol/kg; Guerbet, Paris, France), T1 fast-spin echo with fat saturation (T1 Gd). A detailed scan protocol is provided in Additional file 1. Scoring was carried out according to the RAMRIS for synovitis and BME in the MCP, wrist, and MTP joints and according to Haavardsholm et al. for tenosynovitis in the MCP and wrist [10, 11]. The total MRI inflammation score was calculated by summing the synovitis, BME, and tenosynovitis scores. MRI scoring of the arthritis patients was done independently by two trained readers (WPN and ECN) and the asymptomatic volunteers were scored independently by two trained readers (HWvS and LM). All readers were trained before the start of this project. Readers were blinded for any clinical data. MRI images of asymptomatic volunteers were blinded and mixed with MRI images of RA patients and patients with arthralgia without clinical synovitis (n = 99), to exclude observer bias scoring introduced by knowledge that persons had no symptoms. The within-reader intraclass correlation coefficients (ICC) for the readers who scored the arthritis patients were 0.98 and 0.93, and for the readers who scored the asymptomatic volunteers these were 0.98 and 0.99. The between-reader ICC of the four readers were all above 0.91 (Additional file 2). The mean scores of two readers were used for the analyses.

Associations between MRI-detected inflammation and BMI were assessed using univariable and multivariable linear regression analyses. The MRI inflammation scores were log₁₀-transformed (log₁₀(score + 1)) to approximate a normal distribution. Thereafter, BMI was divided into three categories according to the World Health Organization (WHO) definition: low-normal weight (< 25 kg/m²), overweight (≥ 25 to < 30 kg/m²), and obese (≥ 30 kg/m²). The Kruskal–Wallis test, Mann–Whitney U test, linear regression models, and Spearman rank correlation coefficient were used as appropriate. SPSS V23.0.0 was used.

In RA patients treated in a trial, an association between BMI and BME was observed [3]. We sought to compare our findings in an unselected RA population at disease presentation with these results. In order to do so, we performed a multivariable ordinal logistic regression model in which BME was categorized into quintiles, similar to that done in the trial [3].

In total, 202 RA patients and 170 early arthritis patients with other inflammatory diagnoses were consecutively included in the Leiden EAC. Five RA patients and eight other arthritis patients had no information on height or weight and, respectively, two and three patients had an MRI scan without contrast enhancement and were excluded. Therefore, 195 RA patients and 159 arthritis patients with other inflammatory diagnoses were evaluated. In addition, 196 asymptomatic volunteers were recruited, as already described, three of whom did not receive an MRI scan due to personal problems (n = 1), vasovagal collapse at intravenous puncture (n = 1), and anxiety (n = 1).

Baseline characteristics of all participants are presented in Table 1. Seventy-nine (41 %) RA patients were overweight (BMI > 25 to < 30 kg/m²) and 41 (21 %) were obese (≥ 30 kg/m²); in other arthritis patients these percentages were respectively 44 % (n = 70) and 14 % (n = 23), and in asymptomatic volunteers these percentages were 32 % (n = 61) and 9 % (n = 17) respectively.

In RA patients, a higher BMI was associated with lower MRI inflammation scores (β = 0.97, p = 0.024). A β value of 0.97 indicates that for every point increase in BMI there is a 0.97-fold increase in MRI inflammation score; thus higher BMI was associated with less severe MRI inflammation. In contrast, in other arthritis patients and in asymptomatic volunteers, a higher BMI was associated with higher MRI inflammation scores (respectively β = 1.082, p < 0.001 and β = 1.029, p = 0.040; Fig. 1). When BMI was categorized into three groups (low-normal weight, overweight, and obese), similar results were obtained. Obese RA patients had significantly lower MRI inflammation scores (median = 10.0, IQR = 6.5–18.0) compared with low-normal weight RA patients (median = 19.5, IQR = 7.5–29.0, p = 0.005) and overweight RA patients (median = 14.5, IQR = 7.0–28.0, p = 0.046). Within the group of other early arthritis patients, obese patients had higher total MRI inflammation scores (median = 8.5, IQR = 5.0–13.5) compared with patients with a low-normal weight (median = 2.5, IQR = 0.5–9.5, p = 0.002). Similarly, overweight was also associated with higher total MRI inflammation scores (median = 8.0, IQR = 3.5–22.0, p < 0.001). Within asymptomatic volunteers a tendency towards higher inflammation scores was seen in obese persons (median = 3.5, IQR = 1.0–5.5) compared with low-normal weight persons (median = 1.5, IQR = 0.5–3.5, p = 0.064) and overweight persons (median = 2.0, IQR = 1.0–4.5, p = 0.24; Fig. 1).

Thereafter, the association between BMI (measured continuously) and MRI-detected inflammation was adjusted for age and gender, also showing that a higher BMI was associated with higher inflammation scores in other early arthritis patients (β = 1.036, p = 0.054) and asymptomatic volunteers (β = 1.022, p = 0.040) but with lower inflammation scores in RA patients (β = 0.96, p = 0.005; Table 2 (for nonback-transformed β values, see Additional file 3)). After additional adjustments for CRP and anti-citrullinated protein antibody (ACPA) the results remained similar (β = 0.96, p = 0.003 for RA patients and β = 1.039, p = 0.043 for other early arthritis patients).

The group of early arthritis patients with diagnoses other than RA was divided into the following six subgroups: inflammatory osteoarthritis (n = 38), spondyloarthritis with peripheral arthritis and psoriatic arthritis (n = 40), systemic lupus erythematosus/mixed connective tissue disease and other systemic disease (n = 20), reactive arthritis and lyme arthritis (n = 17), gout and pseudogout (n = 19), and other diseases (n = 25). These subgroups were studied to assess whether the association was more pronounced in a particular disease group. However, the directionality of the effect was similar in all subgroups (Additional file 4).

The total MRI inflammation score is composed of the synovitis, tenosynovitis, and BME scores. To assess whether synovitis, BME, and tenosynovitis have different associations with BMI, these types of inflammation were assessed separately. The synovitis score showed a negative association with BMI in RA patients (β = 0.98, p = 0.047) and a positive association in other arthritis patients and in asymptomatic volunteers (β = 1.084, p < 0.001 and β = 1.031, p = 0.006 respectively). A higher BMI was associated with lower BME in RA patients (β = 0.95, p = 0.002). In other arthritis patients and in asymptomatic volunteers, BMI was not associated with BME scores (respectively β = 1.021, p = 0.24 and β = 1.003, p = 0.79). Within other arthritis patients and in asymptomatic volunteers there was a positive association between BMI and the tenosynovitis score (β = 1.054, p = 0.003 and β = 1.021, p = 0.003 respectively), whereas BMI was not associated with tenosynovitis in RA (β = 0.98, p = 0.21; Additional file 5).

Recently, an inverse association between BMI and BME was shown in RA patients who were treated in a trial [3]. The median BME scores of patients included in this trial and who were low-normal weight, overweight, and obese patients were respectively 9 (IQR 2.5–19), 6.3 (IQR 2.5–13), and 4.8 (IQR 1.5–9.8) [3]. We wished to compare these recent results with our findings obtained in an unselected set of RA patients at the time of disease onset. The median scores observed in our cohort showed a similar tendency but were lower; the median scores in the three groups were respectively 5.0 (IQR 2.0–11.0), 3.0 (IQR 1.0–9.0), and 2.0 (IQR 1.0–4.5). The trial data showed that, after adjusting for race, ACPA, disease duration, DAS, age, and sex, overweight patients had an odds ratio (OR) of 0.68 (95 % CI 0.42–1.08, p = 0.1) and obese patients an OR of 0.47 (95 % CI 0.27–0.82, p = 0.008) for being in a higher BME quintile (Fig. 2). We performed the same analyses without adjusting for race because our study population consisted of 96 % Caucasians, and without disease duration because all of our early arthritis patients were evaluated at their first presentation to the rheumatologic outpatient clinics. In our data, overweight patients had an OR of 0.42 (95 % CI 0.23–0.79, p = 0.007) and obese patients had an OR of 0.30 (95 % CI 0.14–0.64, p = 0.002) compared with low-normal weight patients (Fig. 2). Analyses of both data sets thus showed an inverse association between BMI and BME scores.

The association of BMI with regular measures of inflammation was assessed in RA patients and other arthritis patients. In RA patients the median swollen joint count (SJC) in low-normal weight patients was 5 (IQR = 3–10), in overweight patients the median was 6 (IQR = 2–10), and in obese patients the median was 8 (IQR = 3–10; p = 0.69). The median CRP levels in low-normal weight, overweight, and obese RA patients were respectively 7 mg/L (IQR = 3–23 mg/L), 12 mg/L (IQR = 4–23 mg/L), and 10 mg/L (IQR = 4–24 mg/L; p = 0.39). In other arthritis patients the median of SJC in low-normal weight, overweight, and obese patients were respectively 3 (IQR = 1–5), 3 (IQR = 2–5), and 4 (IQR = 3–7; p = 0.074). Lastly, the median CRP level in the three groups were respectively 4 mg/L (IQR = 3–18 mg/L), 4 mg/L (IQR = 3–12 mg/L), and 5 mg/L (IQR = 3–13 mg/L; p = 0.65). Therefore, in RA patients and in other arthritis patients, BMI was neither associated with swollen SJC nor with CRP levels.

The influence of BMI on MRI-detected inflammation in ACPA-positive or ACPA-negative RA patients was assessed separately. Although the subgroups became small (n = 107 and n = 88 respectively), the effect size of the association between BMI and MRI-detected inflammation remained unchanged (respectively β = 0.97, p = 0.071 and β = 0.97, p = 0.12 for ACPA-positive RA and ACPA-negative RA in the univariable analyses and respectively β = 0.97, p = 0.097 and β = 0.96, p = 0.011 for the analyses adjusted for age and gender).

Furthermore, the association between ACPA level and BMI was assessed with the Spearman rank test in all RA patients. ACPA levels were not correlated with BMI (ρ = −0.33, p = 0.65) and did not differ between the three BMI categories (p = 0.38; Additional file 6).