Clinical evaluation of optical spectral transmission imaging for detection of disease activity in rheumatoid arthritis

Objective To investigate the performance and factors of influence of optical spectral transmission (OST) imaging as a new technique for measuring joint inflammation in rheumatoid arthritis (RA). Method OST was performed in 24 RA patients and 37 controls. Mann–Whitney U-test was used to assess differences in OST score between RA patients and controls. Receiver operating characteristics (ROC), linear regression and generalized estimating equations analysis were used to assess the discriminative capability of OST and the association of OST score with clinical disease parameters, ultrasound, radiographic features and cardiovascular risk parameters. Results Median OST score was higher in RA patients than in controls [16.9 (interquartile range 12.77–19.7) vs 12.11 (10.32–14.93)]. At patient level, OST score was moderately associated with ultrasound [beta 0.38 (95% CI 0.16–0.60), p = 0.001] and clinical disease activity [28-joint Disease Activity Score–C-reactive protein beta 0.30 (95% CI 0.04– 0.57), p = 0.024] in RA patients. In controls, male sex, high body mass index, and hypertension were associated with higher OST scores, while these associations were absent in RA. At joint level, the area under the ROC curve for OST score, with ultrasound or clinical swelling as reference, ranged from 0.63 to 0.70. Joint-space narrowing and malalignment were associated with higher OST joint scores, and subchondral sclerosis with lower scores. Conclusion OST provides an objective measure of synovitis and correlates moderately with other examined disease activity assessment tools. Clinical patient characteristics must be considered when interpreting the results.

Rheumatoid arthritis (RA) is a chronic autoimmune disease characterized by joint inflammation, which can lead to progressive joint destruction and deformity.RA treatment aims to decrease joint inflammation, and the response to treatment is typically evaluated by assessing disease activity.Currently, disease activity is most often measured using the Disease Activity Score based on 28-joint count (DAS28), which is a composite score of subjective reporting, clinical examination, and laboratory findings.The DAS28 is known to be influenced by subjective factors unrelated to joint inflammation (e.g.psychological state, fibromyalgia) and its interrater reliability is questionable, as several studies have reported conflicting interrater agreement (1)(2)(3)(4).
Research has shown that imaging techniques such as musculoskeletal ultrasound (US) and magnetic resonance imaging (MRI) have a higher sensitivity for detecting synovitis than clinical examination (5).However, US is largely operator dependent (6,7) and time consuming, and disadvantages of MRI include high costs and a (relative) contraindication for patients with prostheses of ferromagnetic material, cardiac pacemakers, and implantable defibrillators (8).In addition, these imaging techniques have mainly shown their value for diagnostic purposes, and not for monitoring of therapy (9)(10)(11)(12).
Optical spectral transmission (OST) imaging is a new imaging technique to measure joint inflammation in RA patients (13)(14)(15).Scatter and absorption of light have been shown to be different between inflamed joints and healthy joints (13,(16)(17)(18)(19). Possible explanations for this are that infiltration of inflammatory cells clouds the normally transparent synovial fluid and changes the optical properties of the synovial membrane, in combination with neovascularization and blood-flow changes resulting in increased levels of haemoglobin inside the joint, all resulting in decreased transmission of light (Figure 1) (16,18).In addition, it has been shown that vascular reactivity to a temporal blockade of venous outflow is different between healthy and inflamed joints: more blood accumulates at locations of inflammation and blood accumulation happens more rapidly, resulting in altered transmission of light (13,16).
Therefore, the OST machine uses a pressure cuff on the forearm to temporarily block venous return.The device converts the transmission of light through the joints into a single score, using a prediction algorithm that was built with US as reference (15).Advantages of OST are that it is fast and non-operator dependent.
However, it is known that, in addition to joint inflammation, systemic inflammation inherent to RA affects the vascular system (20)(21)(22)(23).This may influence the outcome of OST in RA joints, since OST also uses blood-flow modulation.In addition, studies have shown that erosions and osteophytes also influence light transmission measurements (15,24).The effects of other frequent bony deformations in RA or concomitant osteoarthritis, such as jointspace narrowing (JSN), subchondral sclerosis, and malalignment of the joint, on OST scores are currently unknown.
To date, a limited number of studies on the performance of OST in external cohorts of RA patients has been carried out (14,15,(24)(25)(26), showing moderate to good correlations of OST scores with clinical, US, and laboratory disease activity markers.Nonetheless, the definite value of OST for clinical practice has not been established yet.Therefore, in this explorative study, we investigated the capability of OST to detect synovitis by comparing OST measurements between RA patients and controls, and by comparing OST measurements with US and clinical disease activity.

Study participants
In this multicentre study, 24 RA patients were recruited during 2018 and 2019 from rheumatology outpatient clinics of Reade and Amsterdam UMC, location VUmc, Amsterdam, the Netherlands.Inclusion criteria were age ≥ 18 years and active disease in the hands, defined as two or more swollen hand joints [(proximal) interphalangeal ((P)IP) and metacarpophalangeal (MCP) joints] or at least one swollen wrist joint.Thirtyseven age-and sex-matched individuals without inflammatory disease were recruited via social media and internal hospital advertisements.Special attention was paid to also include controls with hypertension [systolic blood pressure (BP) > 140 mmHg and/or diastolic BP > 90 mmHg and/or currently on anti-hypertensive treatment].For all participants, exclusion criteria were surgery in the wrists or hands in the preceding 3 months or active concomitant inflammatory disease.A subgroup of RA patients (RA-T group; n = 13, 52%) received biological treatment after the first study visit and had a second visit after 1 month.Results of the RA-T group are shown in the Online Supplementary file.
The study complied with the Declaration of Helsinki and was approved by the Slotervaartziekenhuis and Reade medical ethics committee, Amsterdam, the Netherlands.A written informed consent form was obtained from all participants prior to inclusion.

Healthy joint
Inflamed synovial membrane Neovascularization Clouded synovial fluid

OST measurement
OST measurement was performed using the HandScan device (Hemics, Eindhoven, the Netherlands), by a researcher blinded to the clinical examination and US.Once the seated patient had placed both hands into the device, transmission of laser light at 660 and 808 nm through the hands was recorded, before, during, and after inflation of a pressure cuff (50 mmHg) around the forearms.Regions of interest (ROI) for each joint [bilateral MCP1-5, (P)IP1-5, and wrist; 22 per patient] and reference areas were automatically placed by the software of the device, and manually adjusted when necessary.Joints with a prosthesis or failed measurement were manually excluded.The device gives a score ranging from 0 to 3 for each individual joint, where a cut-off value of ≥ 1 is used to represent inflammation in that joint (27), and a total score comprising a summation of the 22 individual joint scores (range 0-66).

Ultrasound
US was performed according to the European League Against Rheumatism (EULAR) standardized procedure guidelines (28), by an experienced sonographer (> 5 years of experience) or a trained researcher (> 1.5 years of experience), blinded to the clinical assessments and OST.The same 22 joints as examined with OST were assessed with US, using an Aloka P75 or Samsung RS80A with 13-18 MHz linear array transducers.Joints were assessed dorsally in the longitudinal and transverse planes, and wrist assessment included radial carpal, intercarpal, and ulnocarpal joints (the highest score was used when different planes had a different score, e.g. when the radial carpal wrist joint had a score of 2, intercarpal had a score of 1 and ulnocarpal had a score of 0, the definite score for the entire wrist was 2).Grey-scale ultrasound (GSUS) and power Doppler ultrasound (PDUS) were scored using a four-grade semiquantitative scale (0 = no, 1 = mild, 2 = moderate, and 3 = severe signs of inflammation), as originally proposed by Szkudlarek et al (29).Total scores for GSUS, PDUS, and GSUS and/or PDUS (GSUS/ PDUS), were calculated by summation of the individual joint US scores (with ranges of 0-66, 0-66, and 0-132, respectively).A joint was considered inflamed when GSUS > 1 and/or PDUS > 0 (30).

Clinical assessments
Clinical disease activity was assessed with the DAS28 and its individual components [swollen joint count including 28 joints (SJC28), tender joint count including 28 joints (TJC28), patient assessment of global health on a visual analogue scale (VAS), and C-reactive protein (CRP)] (31)(32)(33), and erythrocyte sedimentation rate (ESR).Joint examination was performed by a rheumatology nurse or a trained researcher, blinded to the imaging results.In addition, the SJC and TJC of the 22 joints that were assessed with OST were calculated.

Cardiovascular risk parameters
Cardiovascular risk parameters collected at baseline were age, sex, history and family history of cardiovascular disease, body mass index (BMI), systolic BP, diastolic BP, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, and triglycerides.Hypertension was defined as stated in the inclusion criteria.The European Systematic COronary Risk Evaluation (SCORE) for lowrisk countries was used to estimate the 10 year risk of fatal cardiovascular disease (35).For RA patients, the SCORE was multiplied by 1.5, as recommended by EULAR cardiovascular risk management guidelines (36).

Statistical analysis
Data are presented as mean ± sd for normally distributed continuous variables, median with IQR for skewed continuous variables, and frequencies [n (%)] for categorical variables.The CRP of one patient was missing and imputed using a linear regression model based on ESR.Normality of dependent variables (or residuals after regression analysis) was assessed by visual inspection of the histogram and quantile-quantile plot.For regression analyses, linearity and homoscedasticity were assessed by plotting residuals versus predicted values and observed versus predicted values.When assumptions were not met, log transformation was performed or a non-parametric test was used.For all regression analyses, the log-transformed OST score was used as dependent variable.For continuous independent variables, standardized regression coefficients are reported, representing the number of standard deviations that the log-transformed OST score changes when the independent variable changes by 1 sd.For dichotomous independent variables, the relative difference in log-transformed OST score is reported.An overview of the specific statistical tests can be found in the Online Supplementary file.Statistical analyses were performed using SPSS version 24 and Python version 3.7.2(37)(38)(39)(40).

Results
In total, 24 RA patients and 37 controls were included in the study.Baseline characteristics are shown in Table 1.

OST score in RA patients compared to controls
The median OST total score was significantly higher in RA patients compared to controls [RA 16.9 (IQR 12.68-19.72)vs 12.1 (IQR 10.32-14.93),p = 0.0002] (Supplementary Figure S1).OST score compared to US and clinical disease activity at the patient and the joint level In the RA group, 525 joints were successfully imaged at baseline (three joints were excluded: two prostheses and one measurement failure).Example images of OST+, GSUS/PDUS+ joints, and clinical swelling are shown in Supplementary Figure S2.At the patient level, OST score was positively associated with GSUS score and number of joints with positive US signal (GSUS and/or PDUS) (Table 2).In addition, OST score was positively associated with DAS28-CRP, SJC22, and TJC22.No association was found between OST score and ESR, CRP, or PDUS score.At the joint level, the median OST value was 0.7 (IQR 0.4-1.1)and 172 joints (33%) had a score ≥ 1.In total, 180 joints (34%) were GSUS positive, 111 (21%) PDUS positive, and 190 (36%) GSUS/PDUS positive.During clinical examination, 126 joints (24%) were marked as swollen and 128 (24%) as tender.Overlap between OST+, GSUS/PDUS+ joints, and clinical swelling is shown in Figure 2. The three measures had comparable percentages of positives that were negative for the other two measures (false discovery rates: OST 0.35, US 0.30, and SJC 0.28).
Receiver operating characteristics (ROC) analyses with GSUS/PDUS as reference showed that the area under the curve (AUC) for OST joint score ranged from 0.64 to 0.70, with the lowest score for the wrist and the highest for MCP joints (Figure 2).With clinical swelling as reference, AUCs were slightly lower.ROC curves of GSUS, PDUS, and clinical tenderness were comparable (Supplementary Figure S3).Influence of radiographic features on OST score at the joint level Bony deformations on conventional radiographs were scored in 479 MCP and (P)IP joints (one measurement failure) (Table 3).JSN and malalignment were significantly associated with a higher OST joint score, whereas subchondral sclerosis was associated with a lower OST joint score (Table 3).Osteophytosis was not associated with differences in OST joint score.OST joint scores were higher for joints with erosions or subchondral lucencies, but this did not reach significance.

Influence of cardiovascular risk parameters on OST at the patient level
In controls, male sex, high BMI, high systolic and diastolic BP, hypertension, and high were associated with higher OST measurements at the patient level (Table 4) (see Supplementary Figure S4 for a graphical display).In RA patients, no associations between cardiovascular risk parameters and OST score were found.

Discussion
In this multicentre study, we investigated a new imaging technique to measure joint inflammation in the hands of RA patients.Our main findings are that at the patient level, OST score was higher in RA patients than in controls, and it was moderately associated with US and clinical disease activity.In addition, clinical patient characteristics influencing OST scores were identified, possibly influencing the use of OST in clinical practice.
OST scores were associated more strongly with US than clinical joint swelling, as could be expected since the prediction algorithm was trained on US.Our ROC results are consistent with two studies reporting an AUC of 0.69 and 0.67 with US as reference (24,25), but substantially lower than the studies used to build the HandScan algorithm (AUC 0.81 and 0.85) (14,15).This can be explained by the interobserver variability of US, as the algorithm incorporated in the HandScan was trained on data from only one ultrasonographer, and by the relatively low number of training data on which the algorithm was based (n = 46), which could have led to overfitting (15).We recommend training the algorithm on a new data set with more ultrasonographers and patients, because this will probably improve the performance of the device.We did not find an association between OST score and PDUS.In the study by Triantafyllias et al (24), a smaller association was also found for PDUS than for GSUS (0.442 vs 0.591, respectively).The difference between GSUS and PDUS may be explained by the fact that the OST algorithm was trained on a combination of GSUS and PDUS, or that PDUS is a more difficult technique to master than GSUS, as it is more operator and machine dependent than GSUS [and the algorithm was trained on data from one ultrasonographer on one US machine (14)].
At the joint level, OST joint score, US, and clinical joint swelling had comparable percentages of false positives: positives that were negative for the other two measures.It is unknown whether the OST false positives are truly false positives (noise) or true inflammation that is not detected by US or physical joint examination (higher sensitivity of OST).At the patient level, the difference in OST score between RA patients and controls indicates that OST is able to differentiate between these two groups.Nonetheless, there was considerable overlap, thereby limiting the use of OST as the only measure to detect joint inflammation in the individual patient.In addition, a downside of OST is that it only returns a value between 0 and 3 summarizing all signs of inflammation around the joint, while with US it is possible to separately assess different inflammatory aspects, e.g.tenosynovitis, bone erosions, synovitis, effusion, and enthesitis.For clinical practice, a combination of measures might best be used.A possibility is to also include ESR or CRP, as suggested by a study in which OST score combined with VAS and ESR accurately estimated DAS28-ESR [intraclass correlation coefficient (ICC) internal validation cohort: 0.88 (95% confidence interval 0.87-0.90),and external validation cohort: 0.82 (0.75-0.86)] ( 26) and supported by our results, as we did not find a clear correlation with ESR or CRP, indicating that OST score and ESR/CRP may complement each other.For the use of OST in clinical practice, future research on OST values before and after therapy, and of patients in clinical remission should be performed to determine cut-off values for clinically active disease versus the background signal.In addition, future studies should evaluate OST according to disease activity categories, as OST scores in patients and controls were largely overlapping.
Regarding the influence of radiographic features on OST score at the joint level, we found that JSN and malalignment had higher, and subchondral sclerosis had lower values of OST score, whereas osteophytes did not seem to influence OST score, which is in contrast to previous studies (15,24).Our results indicate a reduced reliability of OST score when bony deformations are present, so caution should be taken when interpreting OST score in these joints.Regarding cardiovascular risk parameters, this study showed that male sex, BMI, and hypertension were associated with higher OST values in controls, but not in RA patients, although they showed a similar trend.This may be caused by joint inflammation obscuring the associations between OST score and cardiovascular parameters, as well as the small sample size.In a larger cohort of RA patients with lower disease activity, associations between OST score and male sex, BMI, and hypertension have been demonstrated (24,41).The higher scores for males may be due to more robust bony, synovial, or tendon structures causing increased light absorption.The higher OST scores for higher BMI may be caused by a possible increase in subcutaneous fat in the hands altering light transmission (24), and the higher scores for hypertension may be caused by altered perfusion.
Fluorescence optical imaging (FOI) is another novel imaging technique, which has both similarities to and differences from OST (42).Both methods use optical imaging to assess inflammation and have a short acquisition time.However, in contrast to OST, FOI requires administration of an intravenous dye (indocyanine green) and evaluation of FOI requires a manual semiquantitative subjective evaluation of the fluorescence signal, which is a disadvantage compared to the automated objective and non-invasive assessment by OST (42).As seen with OST, positive correlations have also been reported between FOI and both US and clinical parameters (Spearman's rho between 0.22 and 0.80) (43).
The present study has some limitations.Because of the small sample size, the results should be interpreted as explorative.For future studies it will be necessary to assess more patients.Owing to the multicentre nature of the study and the breakdown of machinery, US assessments were performed on three different US devices.Nonetheless, except for one patient, all follow-up measurements were performed on the same device.All data were collected once by one investigator, and therefore interobserver and intraobserver agreement could not be investigated.A previous study reported good intraobserver agreement (ICC 0.85, p < 0.001) (15), while interobserver agreement has not been measured yet.In addition, radiographs were not performed at the same time as the OST measurement, but ± 4.5 months before/ after, and new erosive changes may have occurred in between.Nonetheless, it is known that significant changes in structural damage are only visible on X-rays after approximately 2 years.

Conclusion
OST score is an objective measure of synovitis and correlates moderately with other examined disease activity assessment tools.Male sex, BMI, hypertension, and bony deformations were found to influence OST measurements.These factors should be taken into account when interpreting the OST results of the current device in individual patients.To improve reliability, a future version of the device might adjust for these factors when calculating OST scores.

Figure 1 .
Figure 1.Schematic overview of the working principle of optical spectral transmission imaging.Transmission of light is decreased in an inflamed joint (right) compared to a healthy joint (left).

Table 1 .
Descriptive characteristics of study participants.

Table 2 .
Association of optical spectral transmission (OST) imaging with ultrasound and disease activity parameters at the patient level.

Table 3 .
Prevalence of bony deformations on hand X-rays and their relationship to optical spectral transmission imaging (OST) scores at the joint level.

Table 4 .
Association between cardiovascular risk parameters and optical spectral transmission (OST) total score.