The purpose of this study was to determine classification criteria for sympathetic ophthalmia.
Machine learning of cases with sympathetic ophthalmia and 5 other panuveitides.
Cases of panuveitides were collected in an informatics-designed preliminary database, and a final database was constructed of cases achieving supermajority agreement on the diagnosis using formal consensus techniques. Cases were split into a training set and a validation set. Machine learning using multinomial logistic regression was used in the training set to determine a parsimonious set of criteria that minimized the misclassification rate among the panuveitides. The resulting criteria were evaluated in the validation set.
A total of 1,012 cases of panuveitides, including 110 cases of sympathetic ophthalmia, were evaluated by machine learning. The overall accuracy for panuveitides was 96.3% in the training set and 94.0% in the validation set (95% confidence interval: 89.0-96.8). Key criteria for sympathetic ophthalmia included bilateral uveitis with 1) a history of unilateral ocular trauma or surgery and 2) an anterior chamber and vitreous inflammation or a panuveitis with choroidal involvement. The misclassification rates for sympathetic ophthalmia were 4.2% in the training set and 6.7% in the validation set.
The criteria for sympathetic ophthalmia had a low misclassification rate and appeared to perform sufficiently well for use in clinical and translational research.
B Bilateral inflammation after unilateral eye trauma or surgery was first termed sympathetic ophthalmia by Mackenzie in 1840. The ocular inflammation begins weeks to months or even years after an initiating traumatic ocular event, either physical trauma (most often a penetrating ocular injury) or intraocular surgery. The patient then develops bilateral inflammation in both the injured “exciting” eye and in the fellow “sympathizing” eye. Classically, sympathetic ophthalmia was described as a “granulomatous” (ie, with “mutton-fat” keratic precipitates) panuveitis, but with the advent of modern therapy, full-blown disease may not always be seen. Hence, some patients may not have “granulomatous” features and may have minimal anterior chamber inflammation.
Sympathetic ophthalmia is a rare disease which has been declining in incidence. It is estimated to occur in 0.02%-0.05% of cases of ocular trauma and 0.01% of cases of ocular surgery, typically multiple ocular surgeries, particularly vitreoretinal surgery. , A prospective surveillance study in the United Kingdom estimated the incidence as 0.03/100,000 population/year. In that series, ocular surgery was a more frequent cause than traumatic ocular injury. Although nearly all cases occur after penetrating ocular injury or intraocular surgery, sympathetic ophthalmia after trans-scleral laser to the ciliary body, pan-retinal photocoagulation, and radiation therapy for choroidal melanoma have been described, albeit rarely.
Sympathetic ophthalmia is by definition a bilateral uveitis, but observation of inflammation in the exciting eye may be prevented by prior enucleation, phthisis, or corneal opacity. In the era before modern microsurgery and corticosteroid therapy, enucleation of the injured eye typically was performed to prevent sympathetic ophthalmia, and sometimes of the “exciting” eye to improve outcomes in the “sympathizing” eye (a controversial practice), but the low incidence of sympathetic ophthalmia, improvements in globe-preserving surgery, and improvements in therapy largely have led to discontinuation of those practices. Clinical features on ocular examination include anterior chamber inflammation, keratic precipitates, vitreous inflammation, multifocal choroidal infiltrates, and uncommonly serous retinal detachment. The choroidal lesions present as multifocal, small, subretinal yellowish-white spots, and are known histologically as Dalen-Fuchs nodules. In later stages of the disease, the multifocal choroidal lesions become areas of chorioretinal atrophy with loss of retinal pigment epithelium. Those lesions are hyperfluorescent on fluorescein angiography and hypofluorescent on indocyanine green angiography. Similar choroidal lesions can be seen in late-stage Vogt-Koyanagi-Harada (VKH) disease, sometimes termed Dalen-Fuchs-like nodules, and sarcoidosis. Optic disc edema is a recognized complication, and optical coherence tomographic imaging or ultrasonography may demonstrate choroidal thickening.
The histopathology of sympathetic ophthalmia demonstrates an inflammatory infiltrate with mononuclear inflammatory cells (lymphocytes and macrophages) and classically multinucleated giant cells with granuloma formation. Not all cases have granuloma formation, and some cases have only an inflammatory infiltrate of lymphocytes, both T and B cells. Dalen-Fuchs nodules, not found in all cases, are composed of lymphocytes, histiocytes, and de-pigmented retinal epithelial cells. , Expression of HLA-DR can be detected on retinal pigment epithelial cells, leading to speculation about their role in the inflammatory process and as possible antigen presenting cells. However, the pathologic features are similar to other granulomatous eye diseases, such as sarcoidosis.
The Standardization of Uveitis Nomenclature (SUN) Working Group has developed classification criteria for 25 of the most common uveitides by using a formal approach to development and classification. Among the diseases studied was sympathetic ophthalmia.
The SUN Developing Classification Criteria for the Uveitides project proceeded in 4 phases, as previously described: 1) informatics, 2) case collection, 3) case selection, and 4) machine learning ,
As previously described, the consensus-based informatics phase permitted the development of a standardized vocabulary and the development of a standardized, menu-driven hierarchical case collection instrument.
Case Collection and Case Selection
De-identified information was entered into the SUN preliminary database by the 76 contributing investigators for each disease, as previously described. , Cases in the preliminary database were reviewed by committees of 9 investigators for selection into the final database, using the formal consensus techniques described in the accompanying article. , Because the goal was to develop classification criteria, only cases with a supermajority agreement (>75%) that the case was the disease in question were retained in the final database (ie, were “selected”). ,
The final database then was randomly separated into a training set (∼85% of cases) and a validation set (∼15% of cases) for each disease, as described in the accompanying article. Machine learning was used in the training set to determine criteria that minimized misclassification. The criteria then were tested in the validation set; the misclassification rates for both the training set and the validation set were calculated for each disease. The misclassification rate was the proportion of cases classified incorrectly by the machine learning algorithm compared to the consensus diagnosis. For sympathetic ophthalmia, the diseases against which it was evaluated were VKH disease (both early and late stages), Behçet disease uveitis, sarcoidosis-associated panuveitis, syphilitic panuveitis, and tubercular panuveitis.
Comparisons of Subsets of Cases With Sympathetic Ophthalmia
Cases with and without choroidal nodules (“Dalen-Fuchs nodules”) and cases with penetrating ocular trauma versus ocular surgery were compared using the χ 2 test or the Fisher exact test if a cell was <5 for categorical variables and the Wilcoxon rank sum test for continuous variables. P values were nominal and 2-sided.
The study adhered to the principles of the Declaration of Helsinki. Institutional Review Boards (IRBs) at each participating center reviewed and approved the study; the study typically was considered either minimal risk or exempt by the individual IRBs.
A total of 149 cases of sympathetic ophthalmia were collected, and 110 (71%) achieved supermajority agreement for the diagnosis during the “selection” phase and were used in the machine learning phase. Those cases of sympathetic ophthalmia were compared to 806 cases of other uveitides, including 194 cases of Behçet disease, 156 cases of early-stage VKH disease and 103 cases of late-stage VKH, 102 cases of sarcoidosis-associated panuveitis, 70 cases of syphilitic panuveitis, and 277 cases of tubercular panuveitis. Details of the machine learning results for those diseases are outlined in the accompanying article. The characteristics at presentation to a SUN Working Group Investigator of cases with sympathetic ophthalmia are listed in Table 1 . A comparison of cases due to multiple ocular surgeries only versus those due to penetrating ocular injury is presented in Table 2 . Traumatic cases were younger and more often male. There was an apparent shift in the distribution of vitreous cells to higher grades among those with multiple ocular surgeries but no differences in vitreous haze. Cases of sympathetic ophthalmia due to multiple ocular surgeries also were more likely to have exudative detachments and “sunset glow” fundus, although those features occurred in a minority of cases in both subsets. The comparison of cases with and without choroidal lesions are presented in Table 3 . Cases with choroidal lesions were more likely to be chronic and have either no or mutton-fat keratic precipitates. The criteria developed after machine learning for sympathetic ophthalmia are listed in Table 4 . The overall accuracy for panuveitides was 96.3% in the training set and 94.0% in the validation set (95% confidence interval [CI]: 89.0-96.8). The misclassification rates for sympathetic ophthalmia were 4.2% in the training set and 6.7% in the validation set. The disease with which sympathetic ophthalmia most often was confused was tubercular panuveitis.
|Number of cases||110|
|Median IQR (25th, 75th) age, y||43 (25, 59)|
|Asian, Pacific Islander||15|
|Uveitis course, %|
|Keratic precipitates, %|
|Anterior chamber cells, %|
|Anterior chamber flare, %|
|Iris in the sympathizing eye, %|
|Sectoral iris atrophy||0|
|Patchy iris atrophy||0|
|Diffuse iris atrophy||0|
|IOP, involved eyes|
|Median (IQR: 25th, 75th) mm Hg||14 (10, 16)|
|Proportion of patients with IOP >24 mm Hg in either eye, %||4|
|Vitreous cells, %|
|Vitreous haze, %|
|Exudative retinal detachment, %||18|
|Sunset glow fundus, %||10|
|Dalen-Fuchs nodules (multifocal choroiditis), %||63|
|Ocular trauma, %|
|Multiple ocular surgeries||45|
|Penetrating ocular injury||39|
|Penetrating ocular injury followed by multiple ocular surgeries||16|
|Characteristic||Multiple Ocular Surgeries||Penetrating Ocular Injury a||P Value|
|Median IQR (25th 75th) age, y||58 (40, 71)||35 (18, 44)||<.0001|
|Asian, Pacific Islander||20||10|
|Uveitis course, %||.59|
|Keratic precipitates, %||.07|
|Anterior chamber cells, %||.41|
|Anterior chamber flare, %||.51|
|Iris in the sympathizing eye, %||.60|
|Intraocular pressure (IOP), involved eyes|
|Median IQR (25th 75th), mm Hg||14 (9, 16)||14 (10, 16)||.92|
|Percentage of patients with IOP >24 mm Hg in either eye||4||4||1|
|Vitreous cells, %||.01|
|Vitreous haze, %||.37|
|Exudative retinal detachment, %||36||17||.02|
|Sunset glow fundus, %||18||2||.01|
|Dalen-Fuchs nodules (multifocal choroiditis), %||62||63||.94|