Near-infrared Fluorescence Optical Imaging in Early Rheumatoid Arthritis: A Comparison to Magnetic Resonance Imaging and Ultrasonography

DISCUSSION

Near-infrared FOI is a new imaging technology for diagnostic purposes and followup in rheumatic joint diseases. FOI with IV administration of nonspecific fluorophores, such as ICG, depicts disturbances of microvascularization of the examined area. Under inflammatory conditions, vascularization is usually increased compared to healthy tissues, because of highly dysregulated neovascularization^31,32. MRI and US are capable of depicting inflammatory thickening and hypervascularization of the synovial membrane in RA, which is suspected of leading to bone and joint destruction in the course of disease^{17,32,33,34,35,36,37}. In studies with animal models, FOI has shown to be an appropriate imaging method for the detection of arthritis^18,19; however, only a few in vivo studies have been conducted in humans to evaluate FOI as a diagnostic approach in arthritides^{11,13,14,15,17,22,23}. In all these studies, as in this present study, FOI was well tolerated and no adverse events were reported.

To our knowledge, our study is the first to analyze FOI with regard to different joint groups (PIP, MCP, and wrist joints) in comparison to both MRI and US in a cohort of early untreated patients with RA. Because both MRI and US are established imaging methods in early RA, in our study, sensitivity and specificity of FOI under definition of 3 different enhancement phases (P1–P3) were evaluated with MRI as well as GSUS and PDUS as reference examinations.

Highest sensitivity has been shown for P2, especially with regard to the wrist (0.77–0.81) and PIP joints (0.82–1.00), but with unsatisfactory corresponding specificity (0.00–0.41). Sensitivity of P1 and P3 were markedly lower; although their specificity was much higher compared to P2 (Table 2). These results correspond to the findings of 2 other studies that could show that P2 is the most sensitive phase and that P1 and P3 are attributed with higher specificities but lower sensitivities^14,15. Nevertheless, MCP joints constitute an exception in our study: P2 sensitivity (0.43–0.70) is lower than for wrist and PIP joints, but with remarkably higher corresponding specificity (0.77–0.84). Analogous findings have been found in prior studies. Meier, et al compared FOI joint-wise; however, they summarized the carpal joints and MCP joints as a subgroup and compared it to a subgroup consisting of PIP plus DIP joints¹³. Here, sensitivities and specificities were 0.35 and 0.93 for carpal/MCP joints and 0.60 and 0.85 for PIP/DIP joints, respectively. According to analyses by Schafer, et al, FOI is more sensitive with regard to the carpal and PIP joints compared to MCP joints²³; this is confirmed by the results of our study.

The automatically generated composite image (PVM) has shown variable results. PVM has formerly been evaluated in 2 studies showing strong agreement of PVM with MRI as well as clinical examination and moderate to good specificity values; however, sensitivity was moderate in both studies^14,15.

Highest sensitivity values and good agreement rates were found when using PDUS as a reference method. These findings are concordant with the results of 1 study¹⁴ and reflect the fact that PDUS, like FOI (especially P1), primarily evaluates joint hypervascularization regardless of morphological aspects, which are also represented by ICC. GSUS, however, depicts morphological aspects only. As a consequence, correlation to FOI is weak. In contrast, MRI is able to show morphology, hypervascularization, and increased permeability of arterial vessel walls, leading to extravasation of the contrast agent into inflamed areas, while ICG half-life is too short (owing to extensive plasma protein binding and quick metabolization in the liver) to extravasate in sufficient amounts. Higher specificities of P3 (phase after increased signal intensities in the fingertips) indicate the depiction of a certain extravasation by FOI due to increased capillary permeability, which has also been discussed by Werner, et al, who found best specificities and high agreement rates for P1 and P3¹⁵. This investigation is in part supported by the study of Fischer, et al, who reported about arthritic joints that partly show early enhancement and others showing late enhancement¹¹. Again, a study by Meier, et al²² highlighted the importance of early FOI uptake. Indeed, early enhancement of contrast agents is currently widely discussed as an important marker for synovial vascularity in the context of MRI studies^38,39,40,41. On the other hand, Schafer, et al found best results for intermediate (P2) and late enhancement (P3), applying dynamic raw data analysis, although their definition of phases substantially differs from this present study, because phases are defined by time only (P1: 0–120 s; P2: 120–240 s, P3: 240–360 s)²³.

Our study revealed high sensitivity scores of P2 (phase during increased signal intensities in the fingertips); however, these high scores combined with poor specificity and low correlation with MRI and US. This raises the question of whether the discriminatory power of this semiquantitative analysis is high enough to differentiate between physiological enhancement (like in the fingertips) and beginning hypervascularization. The occurrence of false-positive findings, especially in P2, has been discussed: Werner, et al reported normal FOI results in healthy control subjects of 95% to 98%^14,15. Thus, the authors conclude that the disagreement of FOI with MRI and US does not result from false-positive findings, but rather from “subclinical inflammation” not yet detectable by MRI and US^14,15. In a study by Fischer, et al, however, FOI was positive in 8 out of 70 joints in healthy volunteers (11.4%)¹¹.

The dynamic analysis of FOI enhancement has shown to be advantageous with regard to a more standardized and objective image analysis^11,17,22,23. Different approaches have been described so far: 3 studies performed a quantitative analysis of joint enhancement under consideration of the enhancement of the fingertips^11,17,23 and another study compared the rate of early enhancement and areas under the curve for both FOI and MRI, with good to excellent agreement rates also in comparison to the clinical examination²². These studies appear promising compared to the results of some studies that applied a semiquantitative analysis of FOI¹³.

Despite careful planning, this study has some limitations. The geometric arrangement of the camera and light-emitting diodes of the Xiralite system limit the depiction of inflammation at the palmar aspect of the hands. Fluorescence signals may be overlaid by fibrous or muscular structures and fatty tissue, for example with regard to MCP and wrist joints, resulting in a lower detection rate compared to other imaging methods. This is especially true if using cross-sectional imaging methods such as MRI and US, which allow for a more detailed anatomic resolution. Further, FOI unselectively depicts hypervascularized or inflamed areas, so that local hypervascularity in skin lesions, such as small wounds or scratches, may overlay joint structures or even mimic synovitis and lead to false-positive findings because of its nature as a projection imaging technique. Therefore, as mentioned in the methods section, having these conditions was an exclusion criterion for our study. Other known confounders are tendinitis and tenosynovitis of extensor and flexor tendons. In our study, tenosynovitis was scored additionally on MR images as well as US; however, no significant differences were found under inclusion and exclusion of tenosynovitis into the statistical analysis. The method is further limited by ICG because of its high protein binding rate and quick metabolization in the liver^13,17. For better comparability with MRI, dyes with a lower binding rate and a longer half-life would be preferable, especially for better interpretability of the late phase.

Our study has shown variable results owing to 3 phases of FOI enhancement and different analytic aspects. As discussed, comparability to other studies is limited because different analytic approaches have been applied (quantitative vs semiquantitative assessment; variable arrangement of FOI phases; correlation of articular enhancement with enhancement in the fingertips, etc.), a practice that once more illustrates the lack in consistent standards for image analysis and grading of FOI. Further studies should apply clear patient cohorts and should comprise comparative dynamic analyses of MRI and FOI, if possible with histopathological correlation. Another interesting approach would be the analysis of sensitivity and specificity of FOI in combination with clinical and laboratory variables in comparison to MRI as gold standard, which might represent the diagnostic setting in the clinical routine. In addition, longitudinal studies, as performed by Meier, et al²², would allow for the evaluation of FOI with regard to clinical monitoring of patients with arthritis. However, most importantly, larger groups of healthy volunteers need to be examined to clarify the value of FOI for a reliable detection of arthritic joints.

Near-infrared FOI is a fast, nonionizing imaging modality with some potential for application in the diagnostic setting of rheumatic conditions, although this study showed an unreliable sensitivity and specificity of FOI compared to established imaging methods. Nevertheless, our results suggest that the phase-wise approach for a semiquantitative analysis of FOI is necessary and that phase 1 appears to be the most promising for the detection of arthritis although these results partly contradict prior studies. More studies are warranted to elucidate the value of FOI for reliable detection of inflammatory joint disease and for followup assessment of disease activity in patients with rheumatic disease, both in the clinical routine setting and in the context of clinical trials.