Methods

This investigation was conducted with the approval of the Medstar Health Research Institute review board. A retrospective observational case series was performed of 92 eyes of 92 patients with acute, spontaneous fundus-obscuring vitreous hemorrhage. These patients were identified after a review of 4232 charts with the diagnosis code for vitreous hemorrhage (ICD-9 379.23) from January 1, 2003 to October 1, 2013 at a large retina-only subspecialty medical practice. Exclusion criteria included a history of any vascular retinal disease (including diabetic retinopathy), uveitis, or wet macular degeneration in either eye; previous intraocular surgery other than uncomplicated cataract or refractive surgery; and history of recent trauma. Furthermore, patients were excluded if initial presentation was greater than 2 weeks after the onset of symptoms or if follow-up duration was less than 2 months. Finally, patients were excluded if the vitreous hemorrhage was thought to be likely from a cause other than spontaneous PVD, including neovascularization from retinal vein occlusion or retinal arterial macroaneurysm. Inclusion criteria included vision of 20/400 or worse and an ultrasonography scan (B-scan) at the first visit, which indicated that the vitreous hemorrhage was fundus obscuring.

All relevant data from each patient’s progress notes, surgical records, and ancillary testing were abstracted into a standardized data collection sheet that was used in a structured database program for analysis. Patient data recorded from office charts on initial presentation included patient age, sex, involved eye, lens status, the presence or absence of diabetes or hypertension, best-corrected visual acuity (BCVA), and B-scan ultrasound results. The actual B-scan images were not reviewed. Follow-up information recorded included, where applicable, time to spontaneous clearing of the hemorrhage, time to detection of macula-on or macula-off retinal detachment, time to surgical intervention, and indication for surgery. Total follow-up time, final BCVA, and final diagnosis were also noted.

At the initial examination, it was most often concluded that the hemorrhage was likely secondary to a PVD through analysis of the patients’ symptoms, ophthalmic history, examination of the eyes, and ultrasound results. Most patients described a sudden, painless loss of vision in a previously normal eye, with variable reports of increased floaters or photopsias.

To determine the sensitivity and specificity of the preoperative clinical examination, we carefully reviewed the clinic note immediately prior to surgery to determine whether the findings of retinal tears or detachments at the office visit correlated with findings at the time of surgery.

Results

Ninety-two eyes of 92 patients who had dense fundus-obscuring vitreous hemorrhage were included in the study. The series included 48 male and 44 female subjects, ranging in age from 41 to 96 years (mean 63.8). Eight patients had diabetes without any prior retinopathy, 47 had hypertension without any prior retinopathy, and 10 patients were on warfarin. The mean duration of symptoms was 3.9 days prior to presentation (range 0–14 days).

On initial examination the mean initial logMAR BCVA was logMAR 2.218 (Snellen equivalent 20/3000-20/4000, range 20/400-LP). Patients first presented on average 3.9 days after the onset of symptoms (±4.4 days, range 0–14). Fifty-two patients (56.5%) presented within 48 hours of symptom onset, with 25 more patients (27.2%) presenting at 3–7 days and 15 (16.3%) presenting between 1 and 2 weeks. The average duration of follow-up was 490 days with a median of 311 days.

During the course of this study, 72 patients underwent a surgical procedure performed by 1 of 13 surgeons. Six of these patients (1.6%) underwent cryotherapy or laser photocoagulation alone. Sixty-six patients (71.7%) underwent vitrectomy surgery, either alone (n = 62) or in combination with scleral buckling (n = 4). No other surgical procedures were performed as part of the initial management of the presenting condition. On review of operative records, no vision-threatening intraoperative complications were noted in cases where vitrectomy was performed.

Of the eyes that underwent vitrectomy (n = 66), 30 (45.5%) underwent surgical intervention within 7 days of first presentation (early vitrectomy). The 36 others (54.5%) underwent surgery more than 7 days after presentation (delayed vitrectomy). Average time to surgery in the delayed vitrectomy group was 40 days with a median of 29 days. Thirteen eyes were ultimately found to have a diagnosis other than PVD as a cause of the hemorrhage—6 eyes with macroaneurysm, 6 eyes with branch retinal vein occlusion and 1 eyes with central retinal vein occlusion.

Fifty-six of the 92 eyes (60.8%) were discovered to have either a retinal tear or a retinal detachment (RD) (35 RDs and 21 retinal tears). Of the RDs, 9 were macula-off and 26 were macula-on at the time of surgery. With regard to the distribution of macula-on RDs, 16 were found in the early vitrectomy group and 10 were found in the delayed vitrectomy group. All 9 of the macula-off detachments were in the delayed group. The mean time to diagnosis of macula-off RD was 27 days (range 11–60 days). The Figure demonstrates the distribution of days from presentation to the diagnosis of macula-off RD in patients who suffered this complication. Those patients that underwent surgery within the first week of presentation had a significantly lower risk of suffering a macula-off RD ( P < .05).

Time from initial presentation to the diagnosis of macula-off retinal detachment in patients who suffered this complication after developing a dense, fundus-obscuring vitreous hemorrhage. Figure illustrates the time between the initial presentation to our clinic and the discovery of a macula-off rhegmatogenous retinal detachment in each of the 9 patients who developed this complication.

Seventy patients in the study were phakic and 22 were pseudophakic at the time of presentation. Of the 35 patients with RD, 6 were pseudophakic and 29 were phakic at the outset. Lens status did not have a significant effect on the likelihood of developing an RD ( P > .05). At the time of surgery, 10 patients were pseudophakic in the early vitrectomy group and 7 were pseudophakic in the delayed vitrectomy group. Lens status did not significantly correlate with the decision to proceed with early vs delayed vitrectomy ( P > .05). Of the 26 patients who did not undergo vitrectomy, 21 were phakic at the start of the study and 20 remained phakic by study end. In contrast, in the group that underwent vitrectomy, 49 were phakic at the outset and 32 underwent cataract surgery by study end. Undergoing vitrectomy surgery correlated significantly with subsequent cataract surgery in the affected eye ( P < .05).

Seven of the 56 patients (12.5%) with the diagnosis of retinal tear or RD had a history of retinal tear or detachment in the contralateral eye. Of the 36 patients without a retinal tear or detachment, 2 (5%) had a history or retinal tear or detachment in the contralateral eye. There was no statistically significant correlation between having had a retinal tear or detachment in the contralateral eye and having a retinal tear or detachment in the affected eye ( P > .05).

Mean visual acuity improved in the entire cohort from initial presentation (logMAR 2.218, Snellen 20/3000-20/4000) to final follow-up (logMAR 0.318, Snellen 20/40-20/50), and this change was statistically significant ( P < .05). There was no statistically significant difference in visual outcomes among patients that underwent vitrectomy within the first week of presentation (logMAR 0.235, Snellen 20/32-20/40) as compared to all others (logMAR 0.359, Snellen 20/40-20/50, P > .05). Similarly, there was no statistical difference between those who underwent early surgical intervention as compared to those who underwent no surgical intervention (logMAR 0.181, Snellen 20/30, P > .05). However, when looking only at patients who underwent surgical intervention, those who underwent early vitrectomy achieved a significantly better visual outcome as compared to those who did not (logMAR 0.488, Snellen 20/60-20/70; P < .05). Additionally, patients who suffered a macula-involving RD had a significantly worse visual acuity outcome (logMAR 0.722, Snellen 20/100-20/120; P < .05) than all others.

The presurgical examination (including clinical examination and B-scan ultrasonography, where indicated) had a sensitivity of 73.47% and a specificity of 88.24% in determining the presence of retinal tear or detachment. Moreover, the positive predictive value of clinical examination was 94.74% and the negative predictive value was only 53.57%.