This single-site, retrospective cross-sectional study was conducted at the Department of Ophthalmology, Peking Union Medical College Hospital (PUMCH). NIU patients treated with ADA who underwent therapeutic drug monitoring (TDM), including measurements of serum ADA and AAA levels, from June 2024 to April 2025, were included. This study was approved by the PUMCH Ethics Review Committee, and due to its retrospective nature, informed consent was waived.

To minimize variability in ADA levels due to timing of drug administration, only patients who received TDM over 7 days after a scheduled dose of ADA were included, since it takes about 5 to 7 days for ADA to reach peak concentration.

Demographic data including gender and age were obtained by medical record review. Diagnosis, duration of therapy, frequency of ADA administration, coexisting local and systemic treatment, ocular inflammatory activity status, serum ADA, AAA and TNFα level at the time of laboratory examination were documented.

ADA and AAA levels were tested in serum samples using reagents produced by Changde Horui Biotechnology Ltd, China, and analyzers produced by Suzhou Helmen Precision Instruments Ltd, China, based on the principle of fluorescence immunochromatography assay. And TNFα levels were measured using reagents and analyzers produced by Siemens Healthcare Diagnostics Products Ltd, UK, based on the principle of chemiluminescent immunometric assay. The lower and upper limits of quantification for ADA levels are <0.4 μg/mL and >500 μg/mL, respectively; for AAA levels, they are <4 ng/mL and >500 ng/mL, respectively. And for TNFα, the analytical sensitivity is 1.7 pg/mL.

Clinical response was categorized as active or inactive inflammation. Inactive inflammation was determined if the following criteria were all met in both eyes: grade 0 of anterior chamber cells and vitreous haze; resolution or quiescence of previous inflammatory choroidal or retinal lesions; no appearance of new inflammatory choroidal or retinal lesions; and no newly added topical or systemic anti-inflammatory medications at the last visit.

A subset of serum samples from included patients were fractionated by high-performance size-exclusion chromatography (HP-SEC) to characterize ex vivo forms of TNFα. Serum was diluted 1:1 in PBS and filtered (0.22 μm filter) before applying 1.0 mL to a Superdex 200 10/300 GL column (GE Healthcare, UK) and eluted with PBS (0.75 mL/min). Elution profiles were monitored by measuring absorption at 280 nm with an ÄKTA explorer high-performance liquid chromatography system (GE Healthcare, UK). Samples spiked with free TNFα (stored in 6% human serum albumin) [TNFα (500 pg/mL) supplemented with IVIg (5 mg/mL)] and TNFα-ADA complexes [TNFα (500 pg/mL) and ADA (5 μg/mL), supplemented with IVIg (5 mg/mL)] were used as controls (11). Fractions of 0.5 mL were collected and TNFα concentrations were measured using reagents and analyzers mentioned above.

Take P3 generation of THP-1 cells (EallBio Life Sciences, China) and inoculate them in 96-well plates at a density of 1×10⁵ cells per well. Add complete culture medium containing 1, 5, 25, and 125 ng/mL free TNFα and TNFα-ADA complexes respectively. Set 0 ng/mL as the blank control. Each group has 6 duplicate wells. Place them in a CO₂ incubator for 48 hours. Then, collect the supernatant of the culture medium and use the cytokine multiplex detection kits (Cellgene Biotechnology, China) to detect the levels of pro-inflammatory cytokines such as interleukin (IL) -1β, IL-6, IL-8 and interferon γ (IFNγ) in the supernatant.

Statistical analyses were performed using MedCalc, version 23.2.1 (MedCalc Software Ltd, Belgium). Data were expressed as median for continuous variables and frequency (percentage) for qualitative variables. Comparisons of continuous variables were performed using the Mann-Whitney U test. χ² test was used for comparisons involving qualitative variables. A binary logistic regression model was used to generate receiver operating characteristic (ROC) curves to determine threshold ADA level associated with inflammatory activity. Statistical significance was defined as P<0.05.

Notably, median ADA level was lower in patients with AAAs than in those without AAAs (1.9 vs. 6.4 μg/mL; P<0.001). When stratified by clinical response, with or without concomitant immunomodulatory drugs and dosing intervals of ADA, median ADA level remained lower in AAA-positive patients, compared with AAA-negative patients (Table 2).

To investigate the relationship among clinical response, ADA level and presence of AAAs, the cohort was stratified based on active or inactive ocular inflammation. A total of 45 examinations (27.4%) were performed while the inflammation was active. The median ADA level was 2.2 μg/mL in patients with active inflammation, which was significantly lower than 6.0 μg/mL in patients with quiescent inflammation (P<0.001). Furthermore, the proportion of patients with detectable AAAs was higher in the active group (23 of 45 [51.1%]) compared with the inactive group (39 of 119 [32.8%]; P = 0.031). The trend was similar when stratified by coexisting drugs and dosing frequency of ADA, but there was no significant difference in some situations (Table 3).

Analysis of ROC curve demonstrated that an ADA level below the threshold of 4.1 μg/mL was associated with active inflammation, with a sensitivity of 66.7% (95% CI, 51.0%~80.0%) and specificity of 68.1% (95% CI, 58.9%~76.3%) and an area under the curve (AUC) of 0.709, P<0.001 (Figure 1).

The influence of concomitant immunosuppressors on ADA levels and production of AAAs was analyzed. 44 tests (26.8%) were performed when patients were using ADA as systemic monotherapy, while others were performed when patients were taking at least one kind of other immunomodulatory drugs, such as low-dose oral corticosteroids, antimetabolites, calcineurin inhibitors, etc. There was no significant difference in median ADA level between patients with or without concomitant drugs (5.7 vs. 4.0 μg/mL; P = 0.065). But when it comes to the presence of AAAs, the proportion was lower in the antimetabolite (including methotrexate, mycophenolate mofetil and azathioprine) group compared with the ADA monotherapy group (34.3% vs. 54.5%; P = 0.036). On the other hand, calcineurin inhibitors (including cyclosporine and tacrolimus) did not reduce the proportion of AAA positivity (Table 4).

We subsequently investigated the relationship between ADA levels and the time interval since the last administration of ADA. To eliminate the influence of AAAs on ADA levels, we only retained the test results without detectable AAAs and generated a scatter plot (Figure 2). There was significant inter-individual variability in pharmacokinetics of ADA. The median ADA level was 7.0 μg/mL for intervals of ≤14 days, which was significantly higher than the median ADA level of 4.5 μg/mL for longer intervals (P<0.001). However, a considerable proportion of patients with extended dosing intervals still maintained ADA levels above the effective threshold of 4.1 μg/mL, and 90.6% of these patients were under stable control of ocular inflammation.

The correlation among serum TNFα levels, presence of AAAs and clinical response was explored. Analysis demonstrated that median TNFα level was significantly higher in absence of AAAs, and in patients whose ocular inflammation was quiescent (Table 5).

An interesting phenomenon emerged. We traced back 43 results of TNFα level before administration of ADA. It turned out that TNFα levels increased in 42 of 43 patients (97.7%) after ADA therapy. To characterize the forms of TNFα, we analyzed several serum samples of included patients with HP-SEC. We first fractionated standard control samples of free TNFα and TNFα-ADA complexes. TNFα levels were measured in collected fractions. Free TNFα was eluted in the later fractions since the molecular volume was small (Figure 3A), whereas for TNFα-ADA complexes, the TNFα peak shifted to the left (Figure 3B). Next, we characterized serum samples (n=6) with the same protocol and compared the location of TNFα peak with control samples. Two representative graphs of patients are shown (Figures 3C, D). The TNFα peaks of patients’ sera consistently overlapped with control samples of TNFα-ADA complexes, which means after the use of ADA, the majority forms of elevated TNFα were TNFα-ADA complexes.

Subsequently, we compared the differences in pro-inflammatory activity between free TNFα and TNFα-ADA complexes. It was clear that THP-1 cells secreted significantly higher levels of IL-1β, IL-6, IL-8 and IFNγ when they were stimulated with free TNFα, compared with TNFα-ADA complexes (Table 6). It demonstrated that the pro-inflammatory activity of TNFα-ADA complexes was significantly lower than free TNFα.