FormalPara Key Summary Points

The development of a disease activity index could be a major advance in clinical trials as well as in evaluation of treatment response of patients with uveitis in daily clinical practice.

Our group developed a composite disease activity index for patients with uveitis and the objective of the present study was to validate this index.

The study population was patients diagnosed with uveitis and inflammatory activity at the time of their visit to the clinic.

The index classifies inflammatory activity with high discriminatory power.

Introduction

Uveitis is defined as the inflammation of the middle layer of the eye, the uvea. Today, however, uveitis is a generic term used to describe a heterogeneous group of diseases characterized by intraocular inflammation [1]. Uveitis is a major cause of loss of visual acuity and blindness in developed countries [2, 3]. Given the large number of diseases that lead to uveitis, the condition is classified mainly according to the anatomic location (anterior, intermediate, posterior, and panuveitis), although disease course and other clinical and morphological parameters also play relevant roles in defining the disease [4]. In 2005, the Standardization of Uveitis Nomenclature (SUN) Working Group [5, 6] published a series of recommendations with the aim of standardizing the language used to describe and characterize the disease and define its activity. Of note, uveitis that threatens visual acuity affects the posterior pole (intermediate, posterior, and panuveitis). These three anatomic locations often require similar therapeutic strategies—systemic corticosteroids, immunosuppressants, and biologic drugs—and are grouped together to enable clinical studies on treatment, even though their etiologies are different [7, 8].

Studies on the treatment of uveitis are subject to a number of limitations: there are few randomized clinical trials (RCTs), most studies are prospective or retrospective based on clinical experience, and while they show the benefits of treatment, it remains difficult to compare results [9]. The use of immunosuppressants to treat uveitis is widely established in specialized daily clinical practice, although in general there is little evidence to support decisions on therapy.

The main difficulties in performing RCTs and studies on treatment in this field are the lack of high-quality outcome measures and the considerable variability with respect to the choice of outcome measures to evaluate improvement, stability, or worsening of ocular inflammatory activity [10, 11]. This variability implies that none of the instruments used are, in themselves, sufficient to evaluate the course of uveitis; therefore, it is necessary to develop instruments that enable standardized measurement of inflammatory activity, taking into account the main and least expensive parameters applied in daily clinical practice.

Vitreous haze, as defined by Nussenblatt [12], is the reference measure of the United States Food and Drug Administration (FDA) for RCTs evaluating posterior uveitis [13], although the most recent RCTs on the treatment of uveitis with biologics used multicomponent indices to evaluate activity and improvement [14, 15]. Consistent with other authors, we believe that RCTs should not be designed based on a single criterion. Therefore, our group developed a composite disease activity index for patients with uveitis known as UVEDAI [16], which takes into account the overall activity of the eye. The objective of the present study was to validate this index by determining psychometric properties such as construct and criterion validity.

Methods

Design

Multicenter cross-sectional cohort study involving eight multidisciplinary uveitis units (MUUs) in Spanish tertiary hospitals.

Study Population

The study population comprised patients aged ≥ 18 years diagnosed with uveitis involving any anatomical location (anterior, intermediate, posterior, and panuveitis) and inflammatory activity at the time of their visit to the clinic. We excluded patients who were in complete remission, those participating in a clinical trial or an associated research project, and those with postsurgical or traumatic uveitis. The recruitment period ran from June to December 2019. The study was performed at the specialized care level of the National Health Service. We invited eight MUUs from National Health Service hospitals to participate. All of the centers had complete ophthalmological examination equipment, as reported elsewhere [16], which included spectral domain optical coherence tomography (OCT) devices.

Ethics Statement

This study involved human participants and was approved by the Hospital Clínico San Carlos Ethics Committee. Title: Validation and Sensitivity to Change of an Ocular Inflammatory Activity Index: UVEDAI. Internal Code: 18/196-O_SP. Sponsor and Funder: Spanish Society of Rheumatology. Participants signed an informed consent form before participating in the study. The study was performed in accordance with the Declaration of Helsinki.

Study Development

The degree of ocular inflammation was evaluated independently by two ophthalmologists without sharing data so as not to bias the analysis of the index. One ophthalmologist measured inflammatory activity according to daily clinical practice, and the other did so using the UVEDAI score. To minimize variability, the evaluations were conducted by both ophthalmologists during the same visit. In the case of patients with bilateral ocular involvement, the eye selected for the analysis was the most inflamed one.

Operating Definitions and Endpoints

An electronic case report form was designed to collect, in addition to sociodemographic variables (e.g., sex, age, educational level), variables that make up the UVEDAI score (anterior chamber cell grade, macular index, vitreous haze, vasculitis, papillitis, number of chorioretinal lesions, and patient’s global evaluation), anatomical location of uveitis, etiological diagnosis, and evaluation of ocular inflammatory activity according to the physician’s grading, which were stratified as mild, moderate, or high/severe. The operating definitions for the endpoints have been described elsewhere [16].

To reach the objective of validating the UVEDAI, the sample size was estimated at 60 patients. This number was established based on experts’ recommendations who estimated that a sample sized of 50–60 patients is enough to validate an index [17, 18].

Statistical Analysis

Patients were recruited homogeneously and consecutively in the eight MUUs until the sample size was reached. The sample was described, and the distribution of variables was assessed to determine the appropriateness of a parametric analysis. In cases of non-normally distributed variables, the necessary transformations were performed to fit them to Gaussian and/or linear models.

A factor analysis was performed to calculate the construct validity. Given the nature of the variables that make up the UVEDAI (categorical or ordinal), a polychoric correlation matrix was used for entry into the factor analysis [19, 20], since the underlying dimensions thus obtained were much more accurate [21, 22]. Polychoric and tetrachoric correlations (both dichotomous variables) were calculated using FACTOR [23, 24]. Once these matrices were obtained, the corresponding factor analysis was performed; the result was validated using SPSS (IBM SPSS v25). All variables, except for the patient’s global evaluation, had at least five categories. The patient’s evaluation was analyzed by trying various categorization options: five categories (0–1; 2–3; 4–5; 6–7; ≥ 8); four categories (0–2; 3–5; 6–8; ≥ 9); and three categories (< 4 and ≥ 4; < 5 and ≥ 5; < 6 and ≥ 6). Of all the configurations tried, acceptable results were only obtained for the four-category configuration. Construct validity was evaluated using the total variance explained by the components derived from the factor analysis; UVEDAI was considered to have construct validity when the total variance explained on extracting the components was greater than 70%. For purposes of construct validity, the ophthalmologists confirmed that the components extracted corresponded to the units intended to be measured with this index and established a concept for each of the components extracted.

To determine the criterion validity of the UVEDAI, we performed an ordinal multivariate logistic regression analysis, taking into account the ordinal nature of the dependent variable [mild, moderate, and high/severe). The multivariate analysis used to develop the UVEDAI score was re-run, and receiver operating characteristic (ROC) curves were constructed to discriminate between mild and moderate and high/severe uveitis. The analysis was performed using SPSS (IBM SPSS v25).

Results

A total of 62 patients were included in the study [28 men (45%) and 34 women (55%)], with a mean age of 47.7 ± 16.9 years. By diagnosis, 30 (48%) had idiopathic uveitis, 8 (13%) had primary uveitis, 20 (32%) had uveitis associated with systemic disease, and 4 (7%), presented other types of uveitis. By anatomical location, 37 (60%) were anterior, 7 (11%) intermediate, 8 (13%) posterior, and 10 (16%) panuveitis. While a complete examination of both eyes was made in all cases, the eye selected to determine criterion and construct validity was that with greatest activity, that is the one with the most inflammation [32 right (52%), 30 left (48%)]. Information on the variables involved in the UVEDAI score is presented in Table 1.

Table 1 Factors included in the definition of the UVEDAI index
Full size table

Construct Validity

The variables that make up each component, together with their weights, are presented in Table 2. Table 3 presents the number of components extracted and the total variance based on three components, thus explaining 80.32% of the same. Component 1 was identified as anterior chamber involvement, although it includes papillitis, with an inverse correlation (anterior chamber involvement was not detected in patients with papillitis and vice versa). Component 2, posterior pole involvement, included as its most representative variables macular edema, vitreous haze, number of foci of chorioretinal lesions, and vasculitis. Lastly, component 3 includes the patient’s global evaluation, thus explaining 80.32% of the variance in the construct.

Table 2 Matrix of weights of the variables for each of the components according to the number of components extracted
Full size table
Table 3 Total variance explained by the components and by the number of components extracted after Varimax rotation (three components)
Full size table

Criterion Validity

The second ophthalmologist evaluated inflammatory activity in uveitis as mild in 28 cases (45%), moderate in 28 (45%), and high/severe in 6 (10%). Table 4 presents all variables in the UVEDAI used in the logistic regression model and comprising the score that set the cut-offs for classification into three categories of inflammatory activity (mild, moderate, and high/severe).

Table 4 Ordinal logistic regression model of uveitis activity levels (ophthalmologist two), including all variables in UVEDAI
Full size table

The ROC curves for comparing mild uveitis with moderate or high/severe uveitis and high/severe with mild or moderate uveitis are shown in Fig. 1. The area under the ROC curve (AUROC) for the first comparison was 0.867 [95% CI 0.778; 0.955 (p < 0.001)], with a sensitivity of 53.6% for a 10% false-positive rate. In the case of the second ROC curve, which compared uveitis with high/severe activity versus uveitis with mild–moderate activity, the AUROC was 0.946 [95% CI, 0.879; 1.000 (p < 0.001)], with a sensitivity of 83.3% for a 10% false-positive rate.

Fig. 1
figure 1

ROC curves for discriminating between a patients with mild uveitis versus moderate–severe uveitis and b patients with severe uveitis versus mild–moderate uveitis

Full size image

Discussion

Our results confirm the validation of UVEDAI for verifying the construct and criterion validity in this new study population. UVEDAI is a standardized index that provides a valid measurement of global ocular inflammation.

As for construct validity, factor analysis showed that the variance explained with three components was 80.32%, which is an optimal result [25] if compared with the review by Henson and Roberts [26], in which the mean proportion of variance explained by these factors was 52.03%. Two of the three factors removed for analysis corresponded to one anatomical location of uveitis. Anterior chamber cell grade was the variable with the greatest weight for the factor that defined anterior involvement, consistent with the findings of the SUN group [5] for the parameter used to evaluate this involvement. The other variable with greater weight for this component was papillitis, which was only reported in four patients. The correlation between the two variables was high, although it was an inverse correlation; that is, papillitis was not observed in many cases of anterior involvement or vice versa, as observed in clinical practice, since papillitis affects the optic nerve. The dimensions that correlated with component number 2—posterior pole involvement—were vitreous haze, macular edema, number of chorioretinal lesions, and vasculitis, all of which correctly defined this location [5]. In fact, vitreous haze, which was the variable with the greatest weight, is the measurement recommended by the FDA for evaluation of posterior pole activity in RCTs. Factor 3 incorporated the subjective variable, patient’s global evaluation, which is maintained as a relevant item in the index in the same way that it forms part of the other composite indices (DAS28 [27, 28], SDAI and CDAI [29]), which are widely used in the standard examination to evaluate rheumatic disease activity and in RCTs. In our study, each of the variables adequately described each of the three components, and the construct validity was confirmed.

In the absence of a validated gold standard, measurement of criterion validity was not ideal. Similarly, when UVEDAI was applied to the new population, it proved to have good discriminatory and classificatory capacity depending on the degree of inflammatory activity. UVEDAI classified patients with moderate and high activity better than those with mild activity, although both ROC curves showed it to not only have good discriminatory capacity, with values very close to those obtained when it was developed [16], but also the ability to confirm the criterion validity.

The main strength revealed by validation of the present index is its usefulness as a standard tool supported by variables taken from examinations carried out in daily clinical practice, including OCT, without the need for complex diagnostic equipment or invasive tests. Consistent with other experts on validation [30], we believe that the main advantages of composite scores over a set of single measures are avoidance of duplication and greater sensitivity to change. It is noteworthy that UVEDAI, which evaluates global ocular inflammatory activity, not only enables physicians to compare the activity of uveitis independently of the cause, but can also serve as a useful tool for daily patient assessment, evaluation of response to treatment, and performance of future clinical studies and trials with minimal methodological difficulties. Accurate methods that make it possible to evaluate ocular activity in this field must be developed and validated. To our knowledge, no other index that measures this activity has been reported. Another key strength of our study is that, to avoid potential circularity and ensure that the data extracted were reliable, the two ophthalmologists evaluated the patient at the same visit, albeit independently and without sharing the findings of their evaluations.

Our study is limited by the use of the clinical opinion of the ophthalmologist as the gold standard for classifying inflammatory activity. However, as mentioned in the study on the development of UVEDAI [16], our decision was based on the fact that there is currently no reported standard set of criteria to define ocular activity with which we can compare our index. The clinical opinion of the ophthalmologist is habitually used to make clinical and treatment decisions, and some studies define the outcome as an “improvement” in activity according to the physician’s opinion [31]. As pointed out above, the physician’s opinion has been used in the development of other composite indices, such as the Disease Activity Score (DAS) [32] and DAS28 [28], which are widely used both in clinical practice and in RCTs.

Conclusion

The variables included in UVEDAI very accurately describe ocular inflammation. The index also classifies inflammatory activity with high discriminatory power, especially for extreme values. Given that evaluations based on the clinical opinion of the specialist performing the examination are subjective, we believe that having a structured index is highly beneficial when standardizing criteria.

Studies on the interobserver reliability and sensitivity to change of UVEDAI are currently underway to complete its evaluation.