Purpose
To determine risk factors predictive of retinal detachment in patients with cytomegalovirus (CMV) retinitis in a setting with limited access to ophthalmic care.
Design
Case-control study.
Methods
Sixty-four patients with CMV retinitis and retinal detachment were identified from the Ocular Infectious Diseases and Retina Clinics at Chiang Mai University. Three control patients with CMV retinitis but no retinal detachment were selected for each case, matched by calendar date. The medical records of each patient were reviewed, with patient-level and eye-level features recorded for the clinic visit used to match cases and controls, and also for the initial clinic visit at which CMV retinitis was diagnosed. Risk factors for retinal detachment were assessed separately for each of these time points using multivariate conditional logistic regression models that included 1 eye from each patient.
Results
Patients with a retinal detachment were more likely than controls to have low visual acuity (odds ratio [OR], 1.24 per line of worse vision on the logMAR scale; 95% confidence interval [CI], 1.16–1.33) and bilateral disease (OR, 2.12; 95% CI, 0.92–4.90). Features present at the time of the initial diagnosis of CMV retinitis that predicted subsequent retinal detachment included bilateral disease (OR, 2.68; 95% CI, 1.18–6.08) and lesion size (OR, 2.64 per 10% increase in lesion size; 95% CI, 1.41–4.94).
Conclusion
Bilateral CMV retinitis and larger lesion sizes, each of which is a marker of advanced disease, were associated with subsequent retinal detachment. Earlier detection and treatment may reduce the likelihood that patients with CMV retinitis develop a retinal detachment.
Cytomegalovirus (CMV) retinitis is the most common ocular opportunistic infection and the leading cause of blindness among patients with acquired immunodeficiency syndrome (AIDS). Although it is now uncommon in the United States with the widespread use of highly active antiretroviral therapy (HAART), CMV retinitis continues to be the most common ocular opportunistic infection in AIDS patients and an important cause of blindness in many developing countries, especially in Asia. For example, at a tertiary ophthalmology center in Thailand, CMV retinitis affected 33% of AIDS patients and was the second-leading cause of blindness among all patients seen at the clinic. Of the major reasons for visual acuity loss associated with CMV retinitis, retinal detachments are perhaps the most devastating. Patients often have poor long-term visual outcomes even with surgical intervention, leading to permanent loss of vision. Because it can be a vision-threatening complication of CMV retinitis, it is critical to understand the clinical characteristics that are associated with retinal detachment.
Risk factors for retinal detachment have been described for CMV retinitis patients in the United States. However, little information exists on CMV-related retinal detachment in the developing world, where the vast majority of CMV retinitis currently occurs and where retinal detachment is a common complication. The circumstances in such resource-limited settings are very different from those seen in the United States. For example, patients at risk for CMV retinitis in Southeast Asia are not routinely screened, and the disease is often at an advanced stage at the time of initial diagnosis. Patients are often diagnosed with CMV retinitis several months after starting HAART, raising the possibility that retinal infections are present before institution of antiretroviral therapy but not symptomatic until the onset of inflammatory sequelae from immune recovery. inflammation from immune recovery. Furthermore, CMV retinitis in the developing world is treated almost exclusively with intravitreal ganciclovir injections, presenting additional risk for complications. Because the disease course and treatment regimen differ significantly in resource-limited settings, the risk factors for retinal detachment may be different from those found in the United States. In this study, we describe the results from a case-control study conducted in Thailand that investigated the risk factors for CMV retinitis–related retinal detachment.
Methods
Ethical approval was obtained from the Committee on Human Research at the University of California, San Francisco and the Chiang Mai University Faculty of Medicine Research Ethics Committee prior to conducting this case-control study. This study adhered to the tenets of the Declaration of Helsinki.
This study was conducted at the Ocular Infectious Diseases Clinic at Chiang Mai University, a tertiary referral ophthalmology clinic in northern Thailand. The clinic generally examines at-risk HIV patients with visual symptoms. We identified cases of CMV retinitis–related retinal detachment that were diagnosed between December 24, 2004 and August 5, 2010 by reviewing the medical records of all patients with CMV retinitis seen in the Ocular Infectious Diseases Clinic and all patients from the Retina Clinic logbook who had a procedure performed during this time. Cases were defined as any patient for whom retinal detachment was documented in the medical record. We selected 3 controls per case using incidence density sampling. Specifically, we reviewed the Ocular Infectious Diseases Clinic CMV retinitis logbook and enumerated all patients with a visit for CMV retinitis ±3 months from the date the retinal detachment was initially diagnosed. We used a random number generator to select 3 patients without retinal detachment from this list to serve as controls. We sampled controls without replacement; each patient could serve as a control to only 1 case of retinal detachment. If a case had sequential bilateral retinal detachments, the date of the first retinal detachment was used to identify controls.
We collected eye-level and patient-level clinical and demographic information from the visit at which the retinal detachment was first diagnosed for cases, and from the corresponding matched visit for controls. We also collected the same information from the initial clinic visit at which the diagnosis of CMV retinitis was initially made, in order to determine whether features that were present at baseline were predictive of future retinal detachment.
The medical record included detailed retinal drawings for all patients, drawn on a template that included the optic disc, fovea, and vascular arcades, as well as the 3 zones traditionally used for CMV retinitis (zone 1 included the area within 3000 μm of the fovea or 1500 μm of the margin of the optic disc, zone 2 extended from zone 1 to the vortex veins, and zone 3 was the remaining retinal area that extended to the ora serrata). We noted the zone(s) with CMV retinitis and then grouped them based on the most anterior zone that was involved. In addition, we assessed (1) the area of retinitis as a proportion of the retinal surface area, (2) the most anterior extent of the retinitis, expressed as the proportion of the retinal radius as measured from the fovea to the ora serrata, and (3) the most posterior extent of the retinitis, also expressed as the proportion of the retinal radius. Measurements of the size and location of retinitis were performed by 1 of 2 graders using ImageJ software (National Institutes of Health, Bethesda, Maryland, USA), masked to whether the eye was a case or control. Lesion size measurements had good inter-rater reliability (intraclass correlation coefficient [ICC] between the 2 graders on a random set of 50 eyes was 0.88, 95% confidence interval [CI] 0.81–0.94). Based on the charted retinal drawings it was often difficult to clearly distinguish detached retina from areas with CMV retinitis. Therefore, for this study we included lesion size and location only from eyes at the time of initial CMV retinitis diagnosis, before a retinal detachment had developed.
We constructed univariate and multivariate conditional logistic regression models to determine risk factors for retinal detachment, using the eye as the unit of analysis. Separate models were constructed for the matched visit (visit at which the retinal detachment was diagnosed or the time-matched visit for controls) and the visit at which CMV retinitis was first diagnosed. Because eyes from the same person are not independent, we selected only a single eye from each patient for analyses: for cases with sequential bilateral retinal detachments, we selected the first eye that developed a detachment; for cases with bilateral retinal detachments initially diagnosed at the same visit, 1 eye was randomly selected; and for controls with bilateral CMV retinitis, we randomly selected 1 of the eyes. Continuous predictors were tested for specification error (linktest in Stata) and transformations applied as needed. We explored the use of cubic splines for continuous variables, but these did not significantly improve model fit and were therefore not used in the final model. All factors with P < .05 in univariate models were included in the multivariate model, with a backward stepwise selection algorithm employed until all covariates had a P < .10 using the likelihood ratio test. All statistics were performed using Stata version 13 (Statacorp, College Station, Texas, USA).
Results
This study included 64 eyes from 64 patients with CMV retinitis–related retinal detachment and 192 matched control eyes from 192 patients with CMV retinitis but no retinal detachment. Greater than 90% of patients with a retinal detachment were on antiretroviral therapy, and the median CD4 count was 200 cells/μL (interquartile range [IQR] 30–378) at the time of detachment (mean 226 cells/μL, 95% CI 172–279). Most retinal detachments occurred relatively soon after diagnosis of CMV retinitis (median 1.15 months, IQR 0–9.78; mean 8.2 months, 95% CI 4.5–11.9). Visual acuity was generally poor at the time of retinal detachment (median logMAR 1.8, IQR 1.0–1.8 [Snellen equivalent: hand motions, IQR 20/200 to hand motions]; mean logMAR 1.46, 95% CI 1.32–1.60).
Univariate analyses indicated that at the time the retinal detachment was diagnosed, cases with retinal detachment were more likely than controls to have CMV retinitis in the contralateral eye (67.2% vs 52.6%; P = .04), decreased visual acuity (mean logMAR 1.46 vs 0.60 [Snellen equivalent 20/600 vs 20/80]; P < .001), and a shorter period of time since diagnosis of CMV retinitis (mean 8.2 months vs 12.9 months; P = .045; Table 1 ). Patients with retinal detachment were also more likely than controls to have received their initial CMV retinitis diagnosis at the time of the matched visit (35.9% vs 14.1%; P < .001) and therefore less likely to have started management for CMV retinitis before the time of detachment. In multivariate analysis, only decreased visual acuity was strongly associated with retinal detachment (OR 1.24 per line of logMAR acuity, 95% CI 1.16–1.33), although CMV retinitis in the contralateral eye was also included in the model (OR 2.12, 95% CI 0.92–4.90; Table 2 ).
| Characteristic | Initial CMV Retinitis Visit a Mean (95% CI) or N (%) | Retinal Detachment Visit b Mean (95% CI) or N (%) | ||||
|---|---|---|---|---|---|---|
| Cases (N = 41) | Controls (N = 123) | OR (95% CI) | Cases (N = 64) | Controls (N = 192) | OR (95% CI) | |
| Patient characteristics | ||||||
| Age (y) | 37 (34–41) | 36 (31–40) | 1.02 (0.97–1.07) | 38 (34–42) | 38.5 (33–44) | 0.99 (0.96–1.03) |
| Female | 23 (56.1%) | 71 (57.7%) | 0.93 (0.46–1.92) | 30 (46.9%) | 109 (56.8%) | 0.68 (0.39–1.19) |
| CD4 count (cells/μm 3 ) c | 83.0 (37–129) | 78.2 (56–111) | 1.00 (0.98–1.03) | 226 (172–279) | 189 (163–215) | 1.01 (0.99–1.02) |
| Taking antiretroviral therapy d | 26 (92.9%) | 90 (91.8%) | 1.08 (0.16–7.55) | 46 (95.8%) | 163 (95.3%) | 1.24 (0.24–6.33) |
| CMV retinitis in contralateral eye | 26 (63.4%) | 49 (39.8%) | 2.74 (1.25–6.03) j | 43 (67.2%) | 101 (52.6%) | 1.91 (1.03–3.57) j |
| No prior CMV retinitis diagnosis | 41 (100%) | 123 (100%) | – | 23 (35.9%) | 27 (14.1%) | 3.16 (1.67–6.01) j |
| Months since CMV diagnosis e | – | – | – | 8.2 (4.5–11.9) | 12.9 (10.3–15.6) | 0.98 (0.96–1.00) j |
| Eye characteristics | ||||||
| Visual acuity (logMAR) f | 0.87 (0.68–1.06) | 0.72 (0.60–0.80) | 1.04 (0.98–1.10) | 1.46 (1.32–1.60) | 0.60 (0.50–0.70) | 1.24 (1.16–1.33) j |
| Lesion active | 30 (73.2%) | 93 (75.6%) | 0.87 (0.38–2.00) | 18 (28.1%) | 52 (27.1%) | 1.06 (0.55–2.03) |
| Lesion size (% retinal surface) g | 12.8% (9.5–16.1%) | 8.2% (6.7–9.6%) | 2.67 (1.45–4.91) j | – | – | – |
| Most posterior extent of retinitis h | 28.3% (22.1–34.4%) | 27.9% (24.8–31.0%) | 1.01 (0.82–1.25) | – | – | – |
| Most anterior extent of retinitis i | 78.1% (72.9–83.3%) | 72.5% (69.4–75.5%) | 1.26 (0.99–1.60) j | – | – | – |
| Number of lesions | 1.6 (1.3–1.9) | 1.3 (1.2–1.4) | 1.68 (1.03–2.74) j | – | – | – |
| Vitreous haze present | 6 (14.6%) | 6 (4.9%) | 4.37 (1.06–18.11) j | 4 (6.3%) | 3 (1.6%) | 4.00 (0.90–17.87) |
| Retinal hemorrhage present | 18 (43.9%) | 61 (49.6%) | 0.78 (0.37–1.64) | 13 (20.3%) | 43 (22.4%) | 0.89 (0.45–1.75) |
| Frosted branch angiitis present | 6 (14.6%) | 6 (4.9%) | 3.29 (0.99–10.92) | 2 (3.1%) | 1 (0.5%) | 6.00 (0.54–66.17) |
a Longitudinal analysis of the risk factors at the time of initial CMV retinitis diagnosis that predict subsequent retinal detachment; excludes matched groups in which the case had a retinal detachment at the time of initial CMV retinitis diagnosis.
b Cross-sectional analysis of the characteristics present at the time of retinal detachment diagnosis and their association with retinal detachment; includes all matched groups.
c Data not available for all patients. Thirty-two cases (78.0%) and 108 controls (87.8%) had data at the initial visit; 59 cases (92.2%) and 181 controls (94.3%) had data at the matched visit. Odds ratio indicates odds of retinal detachment per 10 cells/μm 3 .
d Data not available for all patients. Twenty-eight cases (68.3%) and 98 controls (79.7%) had data at the initial visit; 48 cases (75.0%) and 171 controls (89.1%) had data at the matched visit.
e Time since diagnosis of CMV retinitis. Numbers in table include all cases and controls; if the 23 cases with a retinal detachment at the time of CMV retinitis diagnosis and their 69 corresponding controls are excluded, the mean time since CMV retinitis diagnosis was 12.8 months (95% CI 7.5–18.1) for cases and 13.4 (95% CI 9.9–16.9) for controls.
f Odds ratio indicates odds of retinal detachment per line of logMAR visual acuity.
g Percent of the total retinal area occupied by the lesion by assessment of retinal drawings; odds ratio indicates odds of retinal detachment per 10% total retinal surface area.
h Distance from the fovea to the most proximal lesion border, calculated as a percentage of the distance from the fovea to the ora serrata; odds ratio indicates odds of retinal detachment per 10% increase in the distance from the fovea to the most posterior border.
i Distance from the fovea to the most distal lesion border, calculated as a percentage of the distance from the fovea to the ora serrata; odds ratio indicates odds of retinal detachment per 10% increase in the distance from the fovea to the most anterior border.
| Characteristic | Initial CMV Retinitis Visit a | Retinal Detachment Visit b |
|---|---|---|
| OR (95% CI) c | OR (95% CI) c | |
| Lesion size, per 10% total retinal surface area | 2.64 (1.41–4.94) | – |
| CMV retinitis in contralateral eye | 2.68 (1.18–6.08) | 2.12 (0.92–4.90) |
| Decreased visual acuity (per logMAR line) | – | 1.24 (1.16–1.33) |
a Longitudinal analysis of the risk factors present at the time of initial CMV retinitis diagnosis that are predictive of subsequent retinal detachment; excludes matched groups in which the case had a retinal detachment at the time of initial CMV retinitis diagnosis.
b Cross-sectional analysis of the characteristics present at the time of retinal detachment diagnosis and their association with retinal detachment; includes all matched groups.
c Initial multivariate models included factors with P < .05 in univariate analyses; the final multivariate models shown above were selected with a backward stepwise algorithm.
In order to assess for risk factors that might predict a subsequent retinal detachment, we performed similar analyses using data from the initial visit at which CMV retinitis was diagnosed. For this analysis, we excluded the 23 cases that had a retinal detachment at the time of initial CMV retinitis diagnosis and their corresponding 69 controls, leaving 41 patients with retinal detachment and 123 controls. In univariate analyses, cases were more likely than controls to have a larger lesion size (mean 12.8% of the total retinal surface area was involved with retinitis, vs 8.2%, P < .001), more peripheral retinitis (the most anterior border of the lesion was on average 78.1% of the total distance from the fovea to the ora serrata, vs 72.5%; P = .049), a higher number of retinal lesions (mean 1.6 vs 1.3; P = .04), CMV retinitis in the contralateral eye (63.4% vs 39.8%, P = .008), and vitreous haze (14.6% vs 4.9%; P = .04; Table 1 ). In the multivariate analysis, only lesion size (OR 2.64 per 10% increase in lesion size measured as a percentage of the total retinal surface area, 95% CI 1.41–4.94) and retinitis in the contralateral eye (OR 2.68, 95% CI 1.18–6.08) remained significant ( Table 2 ).
As noted in the Figure , cases with retinal detachment tended to have more anterior disease at the time of CMV retinitis diagnosis than did controls (46.3% of cases vs 29.3% of controls had disease in zone 3; OR 2.01, 95% CI 0.99–4.09). However, we could not demonstrate an association between zone 3 disease and retinal detachment after adjusting for lesion size (OR 1.54, 95% CI 0.72–3.30).
