The purpose of this study was to determine classification criteria for cytomegalovirus (CMV) retinitis.
Machine learning of cases with CMV retinitis and 4 other infectious posterior/ panuveitides.
Cases of infectious posterior/panuveitides were collected in an informatics-designed preliminary database, and a final database was constructed of cases achieving supermajority agreement on 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 infectious posterior/panuveitides. The resulting criteria were evaluated in the validation set.
A total of 803 cases of infectious posterior/panuveitides, including 211 cases of CMV retinitis, were evaluated by machine learning. Key criteria for CMV retinitis included: 1) necrotizing retinitis with indistinct borders due to numerous small satellites; 2) evidence of immune compromise; and either 3) a characteristic clinical appearance, or 4) positive polymerase chain reaction assay results for CMV from an intraocular specimen. Characteristic appearances for CMV retinitis included: 1) wedge-shaped area of retinitis; 2) hemorrhagic retinitis; or 3) granular retinitis. Overall accuracy for infectious posterior/panuveitides was 92.1% in the training set and 93.3% (95% confidence interval: 88.2-96.3) in the validation set. The misclassification rates for CMV retinitis were 6.9% in the training set and 6.3% in the validation set.
The criteria for CMV retinitis had a low misclassification rate and appeared to perform sufficiently well for use in clinical and translational research.
P rior to the onset of the acquired immunodeficiency syndrome (AIDS) epidemic, cytomegalovirus (CMV) was a rare disease seen primarily in patients undergoing organ transplants, with estimated frequencies of ∼1% for renal transplants and ∼0.5% for bone marrow transplants. The primary risk factor for CMV disease was a CMV serum antibody-positive (“seropositive”) donor organ transplanted into a CMV-seronegative recipient. With the onset of the AIDS epidemic, CMV retinitis became substantially more common, and prior to the widespread use of modern combination antiretroviral therapy (ART) in the mid-1990s, CMV retinitis was among the most common intraocular infections seen in major urban medical centers in the United States and other developed countries. , CMV retinitis is an AIDS-defining opportunistic infection. In that era, the lifetime risk of developing CMV retinitis after the onset of AIDS was estimated at 30%. The primary risk factor was a low CD4 + T-cell count, with a substantial majority of cases occurring among patients with CD4 + T-cell counts ≤50 cells/µL, as CMV-seropositive rates among persons at high risk for human immunodeficiency virus (HIV) infection typically were >90%. With the widespread use of modern ART, the incidence of CMV retinitis among patients with AIDS has decreased by >95%, , primarily due to immune recovery from or prevention of these levels of immunodeficiency and the attendant restoration of immunity to CMV. ,
Cytomegalovirus retinitis occurs in the context of a systemic infection, and CMV can be detected in the blood of most patients with CMV retinitis, either by culture or polymerase chain reaction (PCR) assay of CMV from a blood specimen. , When blood and intraocular specimens from an individual are compared at the time of diagnosis of the retinitis using sequencing of the CMV UL97 gene, there is almost perfect agreement between the 2 isolates. The presumed pathogenesis of CMV retinitis among patients with AIDS is reactivation of latent infection in the context of immunodeficiency, hematologic dissemination to the retina, infection of retinal vascular endothelium, infection of adjacent retina, and unless treated with anti-CMV drugs, spread across the retina. , The end result of CMV retinitis is full-thickness retinal necrosis, leaving a thin, atrophic, and gliotic scar. Retinal detachments were a common complication of CMV retinitis, often due to multiple retinal tears at the border of normal retina and the atrophic scar. The incidence of retinal detachment is related to the extent of retina involved by CMV retinitis, , and its incidence has declined in the era of modern ART. Two clinical morphologic variants of CMV retinitis were described: 1) fulminant or hemorrhagic and 2) granular. The hemorrhagic variant presented with a more extensive area of retinal edema and necrosis, admixed with hemorrhage ( Figure 1 ), whereas the granular variant had a “granular” appearance ( Figure 2 ). The fulminant or hemorrhagic variant has been described as having a “pizza pie” or a “cottage cheese and ketchup” appearance. The only difference observed between the 2 variants was the location of the lesions: the hemorrhagic variant occurred more often in the posterior pole, whereas the granular variant occurred more often in the periphery.
Treatments for CMV retinitis approved by the United States Food and Drug Administration have included intravenous ganciclovir; intravenous foscarnet; intravenous cidofovir; valganciclovir (an oral pro-drug of ganciclovir with good oral bioavailability); the sustained-release intraocular ganciclovir implant; and fomivirsen (an intravitreally administered “anti-sense” drug). With the decline in the incidence of CMV retinitis, production of the ganciclovir implant and of fomivirsen were discontinued. , Intravitreal injections of either ganciclovir or foscarnet have been administered as treatments for CMV retinitis. , However, because of the systemic nature of the CMV infection, absent immune recovery from ART, treatment with intraocular therapy alone is associated with in increased risk of second eye CMV retinitis (among those presenting with unilateral retinitis), visceral CMV disease, and mortality, outcomes still seen the era of modern ART. , Many patients (especially those with lesions threatening the fovea or optic nerve) in developed countries are treated with an initial series of intravitreal injections of either ganciclovir or foscarnet combined with systemic therapy (eg, valganciclovir). With immune recovery from ART, patients with AIDS can recover immunity to CMV and are able to discontinue anti-CMV therapy. , The United States Department of Health and Human Services guidelines include the recommendation to stop anti-CMV therapy when CMV retinitis is inactive and CD4 + T cells have risen to >100 cells/µL for at least 3-6 months. Cohort studies suggest that discontinuation in this circumstance typically can be done safely. The reason for the delay after a rise in CD4 + T cells is that the recovery of immunity to CMV lags the rise in CD4 + T cells by 3-6 months. The rare occurrence of patients with recovered CD4 + T cells but failure to restore immunity to CMV and ongoing problems with recurrent CMV retinitis underscores the importance of immunity specific to CMV for its control.
Patients with CMV retinitis typically have relatively mild amounts of anterior chamber and vitreous inflammation at presentation (reported median vitreous cells grade ½+ and median vitreous haze grade 0+), presumably due to the failure to mount an effective immune response to CMV, and nearly ∼90% will have grade 0 vitritis after treatment with anti-CMV agents. , However, with immune recovery due to ART, the new onset of or an increase in anterior chamber or vitreous inflammation or both can occur, a phenomenon termed immune recovery uveitis (IRU). , Patients with a diagnosis of CMV retinitis after the initiation of ART may have greater levels of anterior chamber and vitreous inflammation than those diagnosed before the initiation of ART as a consequence of IRU.
Even though much of the information on CMV retinitis comes from patients with AIDS, patients with other forms of immunosuppression, such as those undergoing organ transplants and chemotherapy, may develop CMV retinitis. Rare cases have been reported after intravitreal corticosteroid injection or sustained-release corticosteroid implantation, presumably resulting from local, ocular immune compromise. , Cases of CMV retinitis occurring in patients with other types of immune compromise appear to behave similarly to cases of CMV retinitis occurring among patients with AIDS in the modern ART era, including the occurrence of IRU when immunosuppression is reduced or discontinued.
The Standardization of Uveitis Nomenclature (SUN) Working Group is an international collaboration, which has developed classification criteria for 25 of the most common uveitides using a formal approach to development and classification. Among the diseases studied was CMV retinitis.
The SUN Developing Classification Criteria for the Uveitides project proceeded in four 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 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 formal consensus techniques described in the accompanying articles. , 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 the cases) and a validation set (∼15% of the 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. For both the training set and the validation set, the misclassification rate was 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 infectious posterior and panuveitides, the diseases against which CMV retinitis was evaluated were: acute retinal necrosis (ARN), syphilitic uveitis, tubercular uveitis, and toxoplasmic retinitis.
Comparison Between Cases With Aids and Those With Other Immunity-Compromised Conditions
Cases with AIDS were compared to cases with other types of immune compromise (eg, organ transplant, chemotherapy). For categorical variables, a comparison was performed using the χ 2 test or the Fisher exact test when the count of a variable was less than 5. Continuous variables were summarized as medians and compared using the Wilcoxon rank sum test. For characteristics with multiple categorical grades, values above and below the median were compared. P values are nominal and 2-sided.
A total of 251 cases of CMV retinitis were collected, and 211 (84%) achieved supermajority agreement on the diagnosis during the “selection” phase and were used in the machine learning phase. Those cases of CMV retinitis were compared to cases of infectious posterior/panuveitides, including 186 cases of ARN, 174 cases of toxoplasmic retinitis, 35 cases of syphilitic posterior uveitis, and 197 cases of tubercular posterior or pan-uveitis. The details of the machine learning results for these diseases are outlined in the accompanying article. The characteristics of cases with CMV retinitis are listed in Table 1 . The comparison between cases with AIDS and those with other forms of immune compromise is listed in Table 2 . The comparison between cases with AIDS and cases with other forms of immune compromise revealed demographic differences consistent with those of the AIDS epidemic in the United States, and a statistically significant but clinically modest difference in presenting intraocular pressure (median: 13 mm Hg for cases with AIDS; and 15 mm Hg for cases without). In addition, cases with AIDS had less vitritis (median vitreous haze grade 0) than did cases without AIDS (median vitreous haze grade ½+), but the difference was modest. The classification criteria developed after machine learning are listed in Table 3 . Key features of the criteria included necrotizing retinitis with indistinct borders (satellites), immune compromise, and either evidence of intraocular CMV infection (eg, PCR assay results), or a characteristic clinical picture. Features of a characteristic clinical picture for CMV retinitis included a hemorrhagic ( Figure 1 ), granular appearance ( Figure 2 ), or wedge-shaped ( Figure 3 ) retinitis with no to mild vitritis. The overall accuracy for infectious posterior/panuveitides was 92.1% in the training set and 93.3% (95% confidence interval: 88.2-96.3%) in the validation set. The misclassification rate for CMV retinitis in the training set was 6.9% and 6.3% in the validation set. The diseases with which CMV retinitis was most often confused (ie, misclassified) in the training set were toxoplasmic retinitis and syphilitic retinitis and toxoplasmic retinitis and ARN in the validation set.
|Median IQR (25th 75th) age, y||40 (33, 47)|
|Asian, Pacific Islander||9|
|Uveitis course, %|
|Keratic precipitates, %|
|Anterior chamber cells, %|
|Anterior chamber flare, %|
|Iris atrophy (sectoral, patchy, or diffuse)||0|
|IOP, involved eyes|
|Median IQR (25th, 75th) mm Hg||13 (12, 16)|
|Proportion patients with IOP >24 mm Hg in either eye, %||2|
|Vitreous cells, %|
|Vitreous haze, %|
|Number of lesions, %|
|Unifocal (1 lesion)||23|
|Paucifocal (2–4 distinct lesions)||69|
|Multifocal (≥5 distinct lesions)||8|
|Lesion shape, %|
|Round or ovoid||11|
|Placoid or ameboid||35|
|Lesion character, %|
|Lesion location, %|
|Posterior pole involved||38|
|Mid-periphery and periphery only||62|
|Lesion size, %|
|Other features, %|
|Retinal vascular sheathing||36|
|Immunocompromised patients, %|
|Chemotherapy or other immunosuppression||16|
|Laboratory data, %|
|Aqueous or vitreous specimen, PCR-positive for cytomegalovirus||34|