A total of 68 eyes of 68 patients with glaucoma were included. The characteristics of the patients are shown in Table 1. The median (interquartile range, IQR) interval from initial TLE to SBR was 2 months (2.0–3.0).

Figure 2 shows Kaplan–Meier survival analysis. The success rates for QS at 1, 2, and 3 years after surgery were 48.4, 46.8, and 45.1%, respectively. The success rates for CS at 1, 2, and 3 years after surgery were 14.4, 11.2 and 9.6%, respectively. Cox proportional-hazards regression using stepwise method including age, pre-TLE IOP, pre-SBR IOP, interval between TLE and SBR, number of glaucoma medication before TLE, and baseline MD as covariates revealed that only the number of glaucoma medication used before TLE contributed to the survival prediction. An increase of one glaucoma medication increased the hazard ratio for surgical failure by 1.43-fold (95% confidence interval 1.07 to 1.92, P = 0.018). Figure 3 shows Kaplan Meier survival curves by the number of glaucoma medications before TLE. The 3-year QS rate for the patients who used 2, 3, 4, and 5 glaucoma medications before TLE was 71.4%, 60.0%, 55.0%, and 14.3%, respectively. There were 8 patients with 6 or more glaucoma medications prior to TLE, seven of them failed within one year after SBR, and only one survived at 3 years after SBR. The 3-year QS rate for the patients who used 4 or less glaucoma medications before TLE (n = 46) was 58.5%, while that for the patients with 5 or more glaucoma medications before TLE (n = 22) was 16.4% (P = 0.001, Logrank test).

Kaplan–Meier survival curves for the three-year success rate of surgical bleb revision (SBR). The solid black line represents qualified success (QS), and the dotted line represents complete success (CS). The QS rates at 1, 2, and 3 years after surgery were 48.4%, 46.8%, and 45.1%, respectively. The CS rates at 1, 2, and 3 years after surgery were 14.4%, 11.2% and 9.6%, respectively

Kaplan–Meier survival curves for QS rate by the numbers of glaucoma medications used before initial trabeculectomy. The blue line shows the survival curve for patients with two glaucoma medications (n = 7), the green line for those with three (n = 10), the red line for those with four (n = 29), the purple line for those with five (n = 14), the orange line for those with six (n = 4), and the dotted black line for those with seven (4). The 3-year QS rates were 71.4%, 60.0%, 55.5%, 13.4%, and 25.0% for the eye with 2, 3, 4, 5, and 6 glaucoma medications used before TLE. No eye with seven glaucoma medications survived up to 3 years after SBR

Thirty-seven of the 68 included eyes failed within 3 years of surgery; 22 of the 37 failed eyes underwent additional glaucoma surgery, including 9 eyes with repeat TLE, 7 eyes with Ahmed glaucoma valve, 4 eyes with Baerveldt glaucoma implant, one eye with laser cyclophotocoagulation, and 1 eye with laser cyclophotocoagulation followed by Ahmed glaucoma valve. Data from 15 patients who failed within 3 years but did not undergo additional surgery are presented in Supplementary Table 1. In most cases, IOP eventually decreased and settled with eye drops and massage, and surgery was not required. There were five cases of progressive visual field loss, but in three cases the patients did not want additional surgery or did not return. In another case, surgery was not performed because treatment of the contralateral eye took precedence. In the other case, additional surgery was not performed because the eye had developed bullous keratopathy. Excluding these five cases, 41 patients (60.3%) were spared from additional glaucoma surgery for 3 years after SBR.

Figures 4 and 5 show changes over time in IOP and the number of glaucoma medications used. Pre SBR IOP was significantly lower than IOP before TLE (P < 0.001, Wilcoxon test), and IOP decreased significantly after SBR (P < 0.0001, Mixed effect model). At the time of SBR, most patients were using post-TLE eye drops (steroids and antibiotics) and a few were using glaucoma medications. However, glaucoma eye drops were reintroduced early postoperatively after SBR, and the number increased with postoperative time (P < 0.001, Mixed effect model).

Box-and-whisker plots showing changes in IOP over time. Pre-SBR IOP was significantly lower than pre-TLE IOP (P = 0.0001, Wilcoxon test), and IOP decreased significantly after SBR (P < 0.0001, mixed effect model). IOP intraocular pressure, SBR surgical bleb revision, TLE trabeculectomy

Box-and-whisker plots showing changes in the number of glaucoma medication used overtime. The number of medications used before SBR was significantly lower than that before TLE (P < 0.0001, Wilcoxon test). Glaucoma medications were reintroduced early postoperatively after SBR, and their number increased with postoperative time (P < 0.001, Mixed effect model)

Table 2 compares pre- and postoperative IOP, number of glaucoma medications, MD, BCVA, and ECD. Thirty-one eyes that achieved QS at 3 years after SBR were analyzed. The median (IQR) IOP before SBR in all included patients was 21.0 (18.0–25.5) mmHg and was significantly lower than the pre-TLE value, which was 25.5 (21.0–34.0) mmHg (Wilcoxon test, P < 0.0001). However, since in this study SBR is a series of treatments with TLE, the pre-TLE IOP was used as the baseline. Similarly, pre-TLE data were used as the baseline for the number of glaucoma medication, MD, BCVA, and ECD. Because of the short interval between TLE and SBR, some patients had no data on MD, BCVA, and ECD values at the time of SBR. As a result, IOP, the number of glaucoma medications, and ECD significantly decreased from the baseline at 3-year post-SBR. BCVA also decreased postoperatively, but MD did not change from the baseline. Note that both preoperative and postoperative data for MD were available for only 15 eyes.

There were no serious complications during or after SBR. There was no loss of light perception. Complications seen were, transient hypotony with choroidal detachment in 14 eyes, temporary decrease in visual acuity because of hemorrhage spread to the vitreous cavity in 2 eyes, and vitreous opacity due to underlying uveitis in one eye. There were 15 eyes with bleb leak after SBR, but all recovered soon after with either eye ointment or additional suture or both. Shallowing of anterior chamber was seen in 2 eyes but resolved within a month. One eye with uveitic glaucoma developed bullous keratopathy 13 months after SBR (Supplemental Table 1). In this eye, corneal endothelial cells were already reduced before TLE.

Age, type of glaucoma (proportion of POAG), intervals between TLE and SBR, IOP before TLE and SBR, number of glaucoma medications used before TLE, pre-TLE BCVA, and pre-TLE MD were compared between the QS group (n = 31) and the failed group (n = 37) (Table 3). The number of glaucoma medications used before TLE was significantly larger in the failed group than in the QS group (P < 0.01, Mann Whitney U test). There was no significant difference between the two groups in other parameters.