Proliferative diabetic retinopathy (PDR) affects 7% of diabetic patients globally (approximately 17 million people), and half of these patients will be blind if untreated.1,2 However, timely evaluation and treatment can reduce blindness by more than 90%, provide sustained improvement in visual function, improve quality and/or length of life and be highly cost-effective.2–6 Treatment often involves a vitrectomy for non-clearing vitreous haemorrhage and/or tractional retinal detachment involving or threatening the macula.7

Patients with PDR not only suffer from a serious ocular disease but may also have important systemic comorbidities. This is reflected in studies reporting five-year survival following diabetic vitrectomy in predominantly Caucasian populations of between 68–86%, significantly lower than those without PDR (66–99%).8–10 These studies date back two or more decades; survival rates are anticipated to be higher following advances in management of diabetes in the 21st century.

The purpose of this study was to determine contemporary long-term survival rates and associated prognostic factors in patients undergoing diabetic vitrectomy. Survival rates are crucial for decision-making in the clinical setting and for planning healthcare resource allocation. A comparison with earlier studies may also highlight advances and shortcomings in diabetes and more specifically diabetic retinopathy management.

Methods

A retrospective review of clinical records was performed of all patients that underwent vitrectomy for diabetic vitreous haemorrhage and/or tractional retinal detachment. All surgeries were performed between March 2000 and December 2010 at Waikato Public Hospital by two vitreo-retinal surgeons. Morbidity data and survival rates were taken from the time of the initial vitrectomy surgery. Institutional approval was granted from Waikato DHB.

Clinical records were reviewed and data collected on patient demographics, medical comorbidities, serum creatinine, serum HbA1c, medications, pre-operative ocular characteristics and visual acuity. Date of patient death was confirmed from the national health record database.

Statistical analyses were performed using SPSS version 22 (Statistical Package for the Social Sciences GmbH Software, Munich, Germany). P values <0.05 were considered statistically significant. Kaplan-Meier survival curves were analysed for the group. Univariate Cox-regression analyses were performed for all pre-operative variables to identify risk factors for mortality. Significant variables were included in the multivariate analysis.

Results

A total of 182 eyes were included in this study. The mean age was 55 years ± 12 SD (range 22 to 85), 53.8% were male, ethnicity was 49.5% Māori and 40.1% New Zealand (NZ) European (Table 1). Non-Europeans were younger at presentation than NZ Europeans (mean age 52 ± 11 SD vs 59 ± 13 SD, P<0.001) and with a shorter duration of diabetes than NZ Europeans (14 vs 22 years, p<0.001).

Table 1: Ethnic distribution of study population (N=182 patients).

‍

†2013 New Zealand Census.11

Type 1 diabetes mellitus accounted for 16.5%, and type 2 accounted for 83.5% of cases. Mean time from diagnosis of diabetes to time of surgery was 18 years ± 10 (range 0–59). Table 2 displays the medical and medication history of all patients.

Table 2: Medical and medication history as self-reported at the pre-operative assessment or documented in clinical records prior to vitrectomy (N=182 patients).

‍

†Normal creatinine in adult males 60–105 µmol/L, adult females 45–90 µmol/L.
‡Microvascular complication risk increases markedly when HbA1c above 55 mmol/mol.12
§Result of most recent blood test, taken within three months prior to vitrectomy.

Table 3 shows the indications for surgery and pre-operative eye status. Pre-operative best-corrected visual acuity (BCVA) in the operated eye was 1.23 logMAR (~6/120 Snellen equivalent). Post-operative BCVA was 0.471 logMAR (~6/18 Snellen equivalent).

Table 3: Indications for vitrectomy and pre-operative eye characteristics (N=182 eyes).

‍

†Anti-VEGF = anti-vascular endothelial growth factor.

Use of anti-vascular endothelial growth factor (anti-VEGF) therapy in PDR

Anti-VEGF agents were introduced in 2007 at this institution. They were used alone or in conjunction with retinal photocoagulation and/or vitrectomy for treatment of PDR. Following its introduction, the proportion of vitrectomy cases for vitreous haemorrhage reduced from 86.4% to 68.4% (p=0.006). Concurrent intra-vitreal anti-VEGF treatment was given in 23 cases at time of vitrectomy.

Mortality following vitrectomy and associated risk factors

Mean follow-up after vitrectomy was 97 months ± 4 SE (95% CI 88–105, range 2–165) (Figure 1). Ninety patients (49.5%) were deceased as of 31st December 2013. Mean time to death was 60 months ± 4 SE (95% CI 52–68, range 2–158). Following vitrectomy, the three-year survival rate was 83.5% (152/182), and the five-year survival rate was 70.1% (108/154).

Figure 1: Kaplan-Meier survival curve for patients undergoing vitrectomy for proliferative diabetic retinopathy (N=182).

Variables that were significantly associated with mortality on univariate regression analysis were age, non-European ethnicity, pseudophakia, antiplatelet or anticoagulant therapy, dyslipidaemia, dialysis, serum creatinine, neuropathy, hypertension, type 1 diabetes mellitus (Table 4).

Table 4: Univariate Cox proportional hazard model for baseline characteristics associated with mortality in patients undergoing diabetic vitrectomy (N=182).

‍

†VEGF = vascular endothelial growth factor.
‡Result of most recent blood test, taken within three months prior to vitrectomy.
*Statistically significant at p<0.05 level.

Variables that were significantly associated with mortality on stepwise multivariate regression analysis were age, dialysis, and serum creatinine (Table 5).

Table 5: Multivariate Cox proportional hazard model for baseline patient characteristics affecting mortality in undergoing diabetic vitrectomy (N=182).

‍

†Result of most recent blood test, taken within three months prior to vitrectomy.
*Statistically significant at p<0.05 level.

Discussion

Patients with PDR are expected and observed to have much lower survival rates than those without proliferative retinopathy or diabetes, due to severe micro- and macro-vascular complications.8,9,13,14 In the current study, the survival rate following diabetic vitrectomy was 83.5% at three years, reducing to 70.1% at five years. These rates are comparable to previous reports (68–95.8% at five years), suggesting no great change has occurred over the last decade or more.8,9,13,15–17

Diabetes care is costly, and diabetic retinopathy care is a considerable part of this expense.18,19 The costs of admissions for ophthalmic complications at Canterbury DHB, where diabetes was the primary diagnosis, amounted to $579,462 in the 2005/06 financial year, accounting for 18% of total admissions for diabetes.20 Indirect costs, such as those associated with vision loss, are estimated to be three-fold greater than such direct costs.21,22 It is important to continually evaluate the prognosis and survival of these patients in order to guide healthcare management and clinical advisories in the future utilisation of healthcare resources and setting of treatment priorities. The relatively poor survival rates post-vitrectomy need to be considered in the mix with direct and indirect costs both to individual patients and within the wider health economic field.

Despite the high costs associated with diabetic retinopathy, early vitrectomy for diabetic vitreous haemorrhage is associated with a significant gain of 0.41 quality-adjusted life years (QALY), at a relatively low cost per QALY gain (<US$10,000).4,5 Understanding the survival rates and the related prognostic factors may influence clinical decisions around unilateral versus bilateral surgery, complex surgery with poor visual prognosis or high risk of ocular morbidity.

There are reports of exceptional five-year survival rates, as in Japan (95.8%), which are close to that of type 1 diabetics without retinopathy (99%).10,23 This Japanese cohort had a markedly lower mean age (43 years), possibly a shorter duration of diabetes (12.4 ± 7.7 years) and smaller proportion of pre-operative systemic microvascular complications than described in other studies, probably accounting for the high survival rates.23,24 In the current study, half of the patients were Māori, whose life expectancy is, in general, shorter than non-Māori by 7.1 years. Furthermore, with diabetes, the standardised mortality rate for Māori is 62.5 per 100,000 versus 11 per 100,000 for non-Māori.25,26 Thus we would expect a relatively lower survival rate in the current study.

In a New Zealand study by Vote et al, where the cohort was 35% Māori and 26% Pasifika (compared to the current study with 49.5% Māori and 5.5% Pasifika), the mean time to death was 4.3 years following primary vitrectomy for diabetic retinopathy.15 There has been a marginal increase to 5.0 years in the current study, although the overall five-year mortality rate has not changed from around 70% in 1992–1996.15 Between the study by Vote et al and the current study, the mean age at time of surgery has increased from 52 to 55, and the mean time between diagnosis of diabetes and vitrectomy has increased from 16 to 18 years.15,24 This prolonged interval before requiring a vitrectomy may reflect improvements in overall diabetes management over the decade in New Zealand delaying the need for vitrectomy. Although, any effect of better diabetic control on post-vitrectomy survival rates in the later New Zealand cohort may be being offset by the greater age at which vitrectomy is being performed, and thus underestimated.

In the current multivariate Cox regression model, age at time of surgery, dialysis and serum creatinine were statistically significant factors related to reduced survival following diabetic vitrectomy. Matsumato et al reported an association between age, nephropathy and neuropathy with lower survival.23 In the current study, neuropathy was only significant at the univariate level and not in the multivariate model. Similarly, Gollamudi et al showed age, duration of diabetes and nephropathy had an effect on survival in patients undergoing diabetic vitrectomy in the US.8 Markers of nephropathy vary between studies making comparisons difficult. The incidence of diabetes-related end-stage renal disease (ESRD)—the most common cause of ESRD in New Zealand—has doubled between 1992 and 2003 in New Zealand.27 The higher rate of dialysis in the current study as compared with the earlier New Zealand data (14% vs 11%) may reflect this change.15 The similar survival rates, despite worsening nephropathy, may suggest that the management of ESRD in particular has improved.

Māori have an earlier onset of type 2 diabetes mellitus than NZ Europeans (by 8–10 years) and a high failure-to-attend-screening rate (32.3% versus 18.7% overall).28 Thus patients in the current study may be diagnosed late and actually have a longer duration of diabetes before vitrectomy than observed. This may partly account for the lower survival rate in the current study compared to the UK study by Banerjee et al (70% vs 86% at five years), where a longer duration of diabetes was associated with poorer survival.15,24 Although there was no significant association between duration of diabetes and survival in the current study, there is a significant interval before vitrectomy, where diabetic screening and monitoring could be further encouraged, especially for Māori. This may take the form of transport subsidies, integrated specialty services for diabetes to reduce the number of hospital visits and promote a multidisciplinary approach to treatment. Community visits by respective healthcare professionals may also improve awareness of disease and community support for patients.Although pre-operative visual acuity remains around 6/120 (1.23 logMAR in the current study versus 1.38 logMAR a decade earlier in New Zealand), the mean post-operative visual acuity was better [0.471 logMAR (~6/18) versus 1.19 logMAR (~6/90)].15 This improvement may be due to better surgical instrumentation and techniques, more sub-specialty trained vitreoretinal surgeons, improved access to care, less severe disease at time of surgery and more eyes with combined or earlier non-surgical interventions (pan-retinal photocoagulation, long-acting steroids such as triamcinolone and/or anti-VEGF therapy). More than 10% of the current study population had prior anti-VEGF therapy, and 13% had anti-VEGF at time of their diabetic vitrectomy. The significant reduction in vitreous haemorrhage cases since 2007 (when anti-VEGF treatment was introduced for PDR at this institution) may be at least in part due to these new agents. Although the survival rates have not significantly changed over the years, the better visual outcomes following diabetic vitrectomy may result in improved quality of life.

The proportion of vitrectomy patients on insulin has increased from 67% in 1992–1996 to 73.6% despite a reduction in the proportion of type 1 diabetes from 23.7% to 16.5% during the same period.15 This may reflect an increase in appropriate medical treatment of diabetes. Alternatively, the increased use of insulin therapy may reflect a greater disease severity, in which case non-surgical treatments for diabetic retinopathy would appear to be delaying the need for vitrectomy. Furthermore, despite delaying vitrectomy both in terms of diabetes duration and patient age, the proportion of retinal detachments as an indication for diabetic vitrectomy has decreased from 70% in 1992–1996 to 52%, suggesting that patients have less severe ocular disease or present earlier.15 Again, this may be secondary to improvements in screening, medical and laser treatments and access to treatment.

Limitations of this study include its retrospective nature and incomplete data collection. Measures of serum creatinine and HbA1c were within three months of the surgery and were not all taken at pre-operative assessment.

In summary, this study provides updated survival data of patients undergoing diabetic vitrectomy. The relatively poor survival rate associated with the need for diabetic vitrectomy is an important consideration for both individual patient care and for wider health resource allocation. Since the 1990s to 2000s, the mean age of diabetic vitrectomy patients and their duration of diabetes have increased. Furthermore, post-operative visual acuity results have improved. These changes may reflect improvements in eye screening, management of diabetes and retinopathy and improved access to care. The long-term survival of patients undergoing diabetic vitrectomy appears to be associated with age and severe nephropathy. Ongoing efforts are needed to improve the renal care of these patients, especially in non-European populations. Overall, vitrectomy is considered a highly successful and cost-effective intervention, and ongoing updates of survival data are recommended for its utility analysis.

‍