Introduction

Acute retinal necrosis (ARN) is a viral, infectious retinitis characterized by confluent areas of retinal necrosis, vitritis and anterior inflammation. First reported in the 1970s, ARN potentially causes devastating ocular complications and irreversible visual loss in one or both eyes [1]. This destructive uveitis entity occurs most commonly in immunocompetent individuals, but has also been described in immunocompromised hosts. Varicella zoster virus (VZV) and herpes simplex virus (HSV) have both been presumed to be the predominant pathogens leading to ARN [2], while Epstein–Barr virus (EBV) and cytomegalovirus (CMV) have also been implicated [3, 4].

Globally, approaches to treat ARN have focused on intravenous acyclovir (10 mg/kg every 8 h) and oral valacyclovir at disease onset. In recent years, novel antiviral treatments including intravitreal ganciclovir and foscarnet [5] as well as oral valganciclovir or acyclovir have also been considered [6]. Despite these standard therapies, the visual outcome of ARN patients remains poor [7]. Rhegmatogenous retinal detachment (RRD), macula and optic nerve involvement by vasculature destruction have been identified as the main causes of unsatisfactory clinical prognosis, followed by proliferative vitreoretinopathy, macular hole formation and epiretinal membrane [2, 8, 9].

Given the relatively high risk of RRD in ARN, some scholars advocated prophylactic laser retinopexy to reduce the incidence of RRD while several studies opposed [10,11,12,13,14,15,16,17]. However, the design of most previous studies is single-center, retrospective and observational, limiting the reliability and repeatability of conclusion. Taking the rarity of ARN into account, performing multicenter random controlled trials is deemed impracticable [18]. To shed light on this issue, present meta-analysis was conducted to assess the efficacy of prophylactic laser retinopexy in ARN patients.

Methods

The present meta-analysis was conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PRISMA) statement [19].

Search strategy and inclusion criteria

Two independent researchers (SPF and DL) performed the literature search in PubMed, EMBASE databases and Cochrane library from inception to July 2020 using the terms (acute retinal necrosis OR retinal necrosis) AND (laser OR endolaser OR photocoagulation). Reference lists from resulting publications were scanned to avoid omitting relevant citations. Inclusion criteria were as follows: (1) original investigations including retrospective cohort series, interventional and cross-sectional studies, (2) research focused on acute retinal necrosis and prophylactic laser, (3) articles with detailed information regarding baseline characteristics and clinical outcomes. Pre-specified exclusion criteria were as follows: lack of full text manuscript, studies lacking sufficient data, case reports, letters and editorials.

Data extraction and quality assessment

The following data were independently extracted by two investigators (SPF and DL) from each eligible study: leading author name, publication year, sample size, survey period, demographic characteristics of participants, prevalence of RRD and visual outcome at final visit, in accordance with pre-designed inclusion and exclusion criteria. The quality of selected studies was assessed by two independent reviewers (SPF and DL) based on the Newcastle–Ottawa Quality Assessment Scale (NOS). The NOS consists of eight items, divided into three dimensions including selection, comparability and follow-up outcomes. Each item was given one score while the item “comparability of cohorts on the bias of the design or analysis” was given one or two scores according to the baseline characteristics of included studies. If any disagreement occurred, it was resolved through consensus or discussion with a third researcher (YQW). Visual acuity was expressed as logMAR for further comparison and analysis. We converted finger counting, hand motion, light perception and no light perception into logMAR (1.6, 2.0, 2.5 and 3.0, respectively) if the results of best corrected visual acuity (BCVA) were not directly displayed [14].

Statistical analysis

The clinical outcomes including odds ratio (OR) value of RRD incidence and weighted mean difference (WMD) of visual outcome were calculated using random effects model. Cochran’s Q test and I2 test were performed to assess heterogeneity between included studies. In accordance with meta-analysis principle, I2 < 50% indicated low heterogeneity. Subgroup analysis was conducted according to sample size. In addition, sensitivity analysis was used to identify the origin of heterogeneity by sequentially removing one investigation at a time. Egger’s test was adopted to evaluate the publication bias. Meta-analyses were conducted using Stata software (version 12.0, Stata Corp, College Station, TX, USA). Statistical significance was set as p < 0.05 for a two-tailed test.

Results

Study inclusion

In total, 160 citations relevant to our topic were identified from the PubMed (74) and Embase (86) databases. As shown in Fig. 1, after removing 60 duplicated records, 100 studies were screened with title and abstract for eligibility. Consequently, 25 articles were reviewed with full texts. Seventeen publications were removed in present meta-analysis for the following reasons: editorial (n = 1), review (n = 3), without control group (n = 2), laser retinopexy after retinal detachment (n = 1), lacking full text article (n = 2) and lacking data of demographic characteristics and clinical outcomes (n = 8). Eight original articles fulfilling the eligibility criteria were included and subjected to meta-analysis.

Fig. 1
figure 1

Flow diagram of selection process

Full size image

Study demographics

The clinical characteristics of eligible studies are presented in Tables 1, 2 and 3. Eight investigations with 111 prophylactic laser retinopexy in combination with standard antiviral eyes and 136 eyes receiving antiviral treatment alone were retrieved in accordance with the search criteria. Prophylactic laser was performed on the normal retina posterior to necrotic areas, in cases without retinal detachment and vitrectomy. Among these studies, three were performed in American, one was conducted in China, Britain, Korea, Iran and Netherlands, respectively. All studies included in this meta-analysis were designed as retrospective owing to the low occurrence of ARN. If available, detailed demographics including age, gender, interval from onset to diagnosis and extent of retinal necrosis were recorded. The treatment protocol for antiviral and adjuvant therapies varied widely between referral centers. All trials reported the incidence of RRD eyes, whereas visual outcome was not summarized in each study. The methodological quality of most studies ranged from scores of five to six out of a total of nine, except for the investigation conducted by Risseeuw et al. [16], which corrected for the severity of the study cohort’s baseline characteristics (Tab. S1).

Table 1 Characteristics of the included studies
Full size table
Table 2 Demographics of the Included Studies
Full size table
Table 3 Demographics of the included studies
Full size table

Statistical synthesis

There was moderate heterogeneity for RRD risk and a random effects model was selected to synthesize the data. Compared with antiviral treatment alone, the RRD rate decreased in patients receiving prophylactic laser retinopexy, without statistical significance (Fig. 2, P = 0.09, OR = 0.42, 95%CI: 0.15–1.15). In an analysis stratified by sample size, four independent studies demonstrated a notable decrease in RRD risk in the prophylactic laser retinopexy subgroup (P = 0.01, OR = 0.20, 95%CI: 0.06–0.70). No statistical difference was identified among the relatively large sample size subgroups (P = 0.55, OR = 0.64, 95%CI: 0.15–2.75). Three studies confirmed that the prophylactic laser retinopexy subgroup had significant improvement in BCVA during the follow-up period (Fig. 3, P = 0.01, WMD = − 0.98, 95%CI: − 1.74, − 0.22). As shown in Fig. 4, when sequentially excluding one retrospective cohort study to perform sensitivity analysis, the OR value of RRD risk remained stable. This finding confirmed the stability and reliability of this meta-analysis. Moreover, Egger’s test and funnel plots indicated significant publication bias of RRD risk for the eight included studies (P < 0.05).

Fig. 2
figure 2

Forest plot of rhegmatogenous retinal detachment risk across prophylactic laser retinopexy compared with routine antiviral treatment. OR Odd ratio

Full size image
Fig. 3
figure 3

Forest plot of best corrected visual acuity across the prophylactic laser group compared to the routine antiviral group. WMD Weighted mean difference

Full size image
Fig. 4
figure 4

The sensitivity analysis (A) and funnel plot (B) indicated stability and significant publication bias of rhegmatogenous retinal detachment risk, respectively

Full size image

Discussion

ARN is a rare ocular disorder that is found across the world. Generally, ARN results in poor visual outcomes. A systematic review revealed the general RRD rate of ARN is 47% during the follow-up period [20]. RRD occurs in approximately 26–90% of eyes only receiving antiviral treatment and has been reported to be the most common cause of deteriorated visual acuity [2, 12,13,14, 16]. When ARN occurs, the retina turns necrotic approximately two months after disease onset, accelerating the formation of retinal tears and RRD. As previously reported, the median interval between disease onset and RRD is 53 days, but can range from 28 to 165 days [14]. Given this, the use of prophylactic laser retinopexy has been proposed to reduce RRD risk by forming chorioretinal adhesion.

There was significant publication bias in present meta-analysis, as identified by the Egger’s test. Taking various confounding factors into consideration for ARN, observative and small sample investigations with no statistical significance tend not to be submitted or accepted for publication. Moreover, meeting abstracts and dissertations could not be retrieved even if comprehensive and rigorous search was conducted. In addition, it is worthwhile to point out that retrospective design, incomplete clinical features and insufficient data analyses may lead to inevitable publication bias. Thus, the results of this meta-analysis should be interpreted with caution.

The most obvious result emerging from present meta-analysis was no significant advantage of prophylactic laser in preventing RRD. This finding was in agreement with the conclusion of two previous studies, both of which had relatively large sample size [14, 16]. Subgroup analysis by sample size demonstrated significantly reduced RRD risk in studies with small samples, indicating the potential for inaccurate conclusions in single-center studies. Patients undergoing prophylactic laser retinopexy must have relatively mild vitritis to allow clear visualization of peripheral retina. The application of laser is limited in cases with severe vitreous haze, therefore precluding any attempt at laser and leading to selection bias. Severe vitritis predisposed to accompany with wider extent of involved retina, facilitating the formation of retinal tears and increasing RRD risk [14, 21]. A possible explanation for this result is initiating agents for disease progressing remain unsettled. Pro-inflammatory cytokines and chemokines including monokine induced by interferon γ, interferon γ, interferon-γ-inducible protein-10, interleukin-6, interleukin-8 and interleukin-17 have been shown to have essential roles in the pathogenesis of ARN relative to other ocular disorders [22, 23]. The imbalance between pro- and anti-inflammatory reactions contributes to the rapid and destructive development of ARN [24]. Performance of laser retinopexy could not remove the inflammatory mediators and high viral load, which have been considered to be the predominant cause of retinitis and vitritis. The vitreous traction, another crucial factor of RRD, was also not relieved by prophylactic laser. Another noteworthy issue is photocoagulation at the previously necrotic retina may cause retinal damage and promote the formation of atrophic hole, indicating the importance of suitable laser spots [10, 15]. Although significantly improved visual outcome was observed in the prophylactic laser subgroup, only three studies have reported quantifiable results for visual prognosis, which were affected by both selection bias and small study size effects [10, 11, 17]. It is worth mentioning that the possibility of a benefit from such prophylactic effects should be considered if assessed in a multicenter study.

The antiviral regimen and adjuvant treatment varied widely among referral centers, with different treatment patterns likely leading to discrepancies in clinical outcomes. Retrospective investigations demonstrated that intravenous acyclovir significantly decreased the incidence of fellow eye involvement and accelerated disease resolution [25, 26]. Therefore, intravenous acyclovir has been the most widely adopted treatment for ARN worldwide, as described in all included studies. Recent studies recommended that oral valacyclovir or famciclovir effectively improved the visual acuity and decreased the RRD rate in ARN [27, 28]. Intravitreal ganciclovir or foscarnet has also been identified as an efficient option to alleviate inflammation [29, 30]. However, Tibbetts et al. [14] discovered that various antiviral therapies did not affect clinical visual outcome of ARN. Detailed investigations of therapeutic effects and bioavailability of various antiviral treatments will generate advanced insights for ophthalmologists.

This study had several limitations. First, the sample sizes of previous studies were relatively small, owing to low incidence of ARN. In addition, all included studies were retrospective and cross section, limiting the level of evidence owing to selection bias. Multiple studies did not control for disease severity and other influential factors such as the interval from diagnosis to laser retinopexy, follow-up period. Another essential concern is the lack of gold standard of assessment for the extent of involved retina. Various grading criteria including clock hours of involved retina, percent of necrotic areas and cumulative number of quadrants are adopted, limiting the direct comparisons among included studies. As a result, concisions from these studies are likely inconclusive. Standard definitions of treatment success also vary extensively, including RRD rate, visual and anatomic prognosis, and interval from treatment to remission. Further studies that take these variables into account will be needed in the future.

Conclusions

In conclusion, this meta-analysis provides comprehensive and updated elucidation of prophylactic laser retinopexy in ARN. Our results do not support convincing evidence of prophylactic laser in preventing RRD. Taken present observations into consideration, a multicenter investigation with corrected extent of involved retina will be needed to explore the role of prophylactic laser retinopexy and evaluate long-term clinical prognosis.