Tamponade in surgery for retinal detachment associated with proliferative vitreoretinopathy

Stephen G Schwartz; Harry W Flynn Jr; Xue Wang; Ajay E Kuriyan; Samuel A Abariga; Wen-Hsiang Lee

doi:10.1002/14651858.CD006126.pub4

Tamponnement dans l'opération du décollement de la rétine associé à une vitréorétinopathie proliférante

Authors' declarations of interest

Version published: 13 May 2020 Version history

https://doi.org/10.1002/14651858.CD006126.pub4

Collapse all Expand all

Résumé scientifique

available in

Contexte

Le décollement de la rétine (DR) avec vitréorétinopathie proliférative (VRP) nécessite souvent une intervention chirurgicale pour rétablir une anatomie normale et stabiliser ou améliorer la vision. La VRP se produit généralement en association avec un DR récurrent (c'est‐à‐dire après une première chirurgie de rattachement de la rétine), mais elle pourrait parfois être associée à un DR primaire. Dans les deux cas, un agent de tamponnement (gaz ou huile de silicone) est nécessaire pendant l'opération pour réduire le taux de récidive postopératoire.

Objectifs

L'objectif de cet revue était d'évaluer la sécurité et l'efficacité relatives de divers agents de tamponnement utilisés lors de la chirurgie pour le DR compliqué par la VRP.

Stratégie de recherche documentaire

Nous avons effectué une recherche dans CENTRAL (qui contient le registre des essais du groupe Cochrane sur l'ophtalmologie) (The Cochrane Library 2019, numéro 1), Ovid MEDLINE, Ovid MEDLINE In‐Process and Other Non‐Indexed Citations, Ovid MEDLINE Daily, Ovid OLDMEDLINE (de janvier 1946 à novembre 2019), EMBASE (de janvier 1980 à novembre 2019), Latin American and Caribbean Literature on Health Sciences (LILACS) (de janvier 1982 à novembre 2019), le métaRegistre des essais contrôlés (mRCT) (www.controlled‐trials.com), ClinicalTrials.gov (www.clinicaltrials.gov) et le système d'enregistrement international des essais cliniques de l'OMS (ICTRP) (www.who.int/ictrp/search/fr). Nous n'avons appliqué aucune restriction concernant la langue ou la date lors des recherches électroniques d'essais. Nous avons consulté les bases de données électroniques pour la dernière fois le 2 janvier 2019.

Critères de sélection

Nous avons inclus des essais contrôlés randomisés (ECR) sur des participants subissant une intervention chirurgicale pour un DR associée à une VRP, qui comparaient divers agents de tamponnement.

Recueil et analyse des données

Deux auteurs de la revue ont examiné les résultats des recherches de façon indépendante. Nous avons utilisé les procédures méthodologiques standard définies par Cochrane.

Résultats principaux

Nous avons identifié quatre ECR (601 participants) qui ont fourni des données pour les critères de jugement primaires et secondaires. Trois ECR ont fourni des données sur l'acuité visuelle, deux sur l'attachement maculaire, un sur le rattachement rétinien et deux sur des événements indésirables tels que le DR, l'aggravation de l'acuité visuelle et la pression intraoculaire.

Caractéristiques des études

Les caractéristiques des participants variaient selon les études et les groupes d'intervention, avec une fourchette d'âge comprise entre 21 et 89 ans, et une majorité d’hommes. L'étude sur les silicones a été menée aux États‐Unis et a consisté à la réalisation de deux ECR: (l’huile de silicone par rapport à l'hexafluorure de soufre (SF₆) ; 151 participants) et (l’huile de silicone par rapport au perfluropropane (C₃F₈) ; 271 participants). Le troisième ECR a comparé l'huile de silicone lourde (un mélange de perfluorohexyloctane (F₆H₈) et d'huile de silicone) à l'huile de silicone standard (soit 1000 centistokes ou 5000 centistokes ; 94 participants). Le quatrième ECR a comparé 1000 centistokes d’huile de silicone à 5000 centistokes d'huile de silicone chez 85 participants. Nous avons évalué la plupart des ECR avec un risque de biais faible ou incertain pour la plupart des domaines de « risque de biais ».

Résultats

Bien que le gaz SF₆ ait été associé à des critères de jugement anatomiques et visuels plus mauvais que l'huile de silicone à un an (données quantitatives non communiquées), à deux ans, l'huile de silicone n'a pas montré de données probantes indiquant une différence d'acuité visuelle comparée au gaz SF₆ (33 % contre 51 % ; risque relatif (RR) 1,57 ; intervalle de confiance (IC) à 95 % 0,93 à 2,66 ; 1 ECR, 87 participants ; données probantes d’un niveau de confiance faible). À un an, un autre ECR comparant l'huile de silicone et le gaz C₃F₈ n'a pas trouvé des données probantes indiquant une différence d'acuité visuelle entre les deux groupes (41 % contre 39 % ; RR 0,97 ; IC à 95 % 0,73 à 1,31 ; 1 ECR, 264 participants ; données probantes d’un niveau de confiance faible). Dans un troisième ECR, les participants traités avec de l'huile de silicone standard n'ont pas non plus montré de données probantes indiquant une différence dans le changement de l'acuité visuelle à un an, mesuré sur l'échelle logMAR, comparés à ceux recevant de l'huile de silicone lourde (différence moyenne ‐0,03 logMAR ; IC à 95% ‐0,35 à 0,29 ; 1 ECR ; 93 participants ; données probantes d’un niveau de confiance faible). Le quatrième ECR qui comparait l’huile de silicone à 5000 centistokes et celle à 1000 centistokes n'a pas rapporté de données sur l'acuité visuelle.

Pour l'attachement maculaire, les participants traités à l'huile de silicone pourraient probablement connaître des résultats plus favorables que les participants qui ont reçu du SF₆ à la fois à un an (données quantitatives non communiquées) et à deux ans (58 % contre 79 % ; RR 1,37 ; IC à 95 % 1,01 à 1,86 ; 1 ECR ; 87 participants ; données probantes d’un niveau de confiance faible). Un autre ECR comparant l'huile de silicone au C₃F₈ à un an, n’a pas trouvé de données probantes indiquant une différence concernant l'attachement maculaire (RR 1,00 ; IC à 95% 0,86 à 1,15 ; 1 ECR, 264 participants ; données probantes d’un niveau de confiance faible). Un ECR qui a comparé 5000 centistokes à 1000 centistokes a rapporté que le rattachement de la rétine a été un succès chez 67 participants (78,8%) avec la première chirurgie et 79 participants (92,9%) avec la deuxième chirurgie, et n’a pas rapporté de données probantes indiquant une différence entre les groupes (1 ECR ; 85 participants ; données probantes d’un niveau de confiance faible). Le quatrième ECR qui a comparé l'huile de silicone standard à l'huile de silicone lourde n'a rien rapporté concernant l'attachement maculaire.

Événements indésirables

Dans un ECR (86 participants), comparant ceux qui ont reçu de l'huile de silicone standard à 1000 centistokes à ceux qui ont reçu de l'huile de silicone à 5000 centistokes, n’a pas montré de données probantes indiquant un différence concernant l'élévation de la pression intraoculaire à 18 mois (24% contre 22% ; RR 0,90 ; IC à 95% 0.41 à 1,94 ; données probantes d’un niveau de confiance faible), dans une cataracte visuellement significative (49 % contre 64 % ; RR 1,30 ; IC à 95% 0,89 à 1,89 ; données probantes d’un niveau de confiance faible) et dans une incidence de décollement de la rétine après l'élimination de l'huile de silicone (RR 0,36 IC à 95% 0,08 à 1,67 ; données probantes d’un niveau de confiance faible). Un autre ECR comparant l'huile de silicone standard à l'huile de silicone lourde suggère qu’il n’y a pas de différence dans le décollement de la rétine à un an (25% contre 22% ; RR 0,89 ; IC à 95% 0,54 à 1,48 ; 1 ECR ; 186 participants ; données probantes d’un niveau de confiance faible). Le décollement de la rétine n'a pas été signalé dans les ECR qui ont comparé l'huile de silicone au SF₆ et l'huile de silicone au C₃F₈.

Conclusions des auteurs

Il ne semble pas y avoir de différences majeures dans les critères de jugement entre le C₃F₈ et l'huile de silicone. L'huile de silicone pourrait être meilleur que le SF₆ pour l'attachement maculaire et d'autres critères de jugement à court terme. Le choix d'un agent de tamponnement doit être propre à chaque patient. Tant l'utilisation du C₃F₈ que celle de l'huile de silicone standard semble raisonnable pour la plupart des patients atteints de décollement de la rétine associée à la vitréorétinopathie proliférante. L'huile de silicone lourde, qui n'est pas disponible pour un usage clinique de routine aux États‐Unis, pourrait ne pas démontrer de données probantes indiquant une supériorité par rapport à l'huile de silicone standard.

PICOs

Population

Intervention

Comparison

Outcome

The PICO model is widely used and taught in evidence-based health care as a strategy for formulating questions and search strategies and for characterizing clinical studies or meta-analyses. PICO stands for four different potential components of a clinical question: Patient, Population or Problem; Intervention; Comparison; Outcome.

See more on using PICO in the Cochrane Handbook.

Résumé simplifié

available in

Tamponnement dans l'opération du décollement de la rétine associé à une vitréorétinopathie proliférante

Quel est l’objectif de cette revue ?
Le but de cette revue Cochrane était de déterminer si les substances appelées agents de tamponnement utilisées pour traiter le décollement de la rétine (DR) associé à la vitréorétinopathie proliférante (VRP) sont sûres et efficaces. La VRP fait référence à la croissance et à la cicatrisation de la rétine.

Principaux messages

Le choix d'un agent de tamponnement doit être propre à chaque patient. L'utilisation de C₃F₈ (un type de gaz) ou d'huile de silicone standard semble raisonnable. L'huile de silicone lourde, qui n'est pas disponible pour un usage clinique de routine aux États‐Unis, ne présente aucun avantage ou bénéfice par rapport à l'huile de silicone standard.

Qu'étudie cette revue ?
La rétine est le tissu de détection de la lumière le plus profond de l'œil (semblable à la pellicule d'un appareil photo), et son fonctionnement normal dépend de sa fixation à la couche sous‐jacente. Le DR est un trouble de l'œil dans lequel la rétine se sépare physiquement de la couche de tissu sous‐jacente. La macula est la partie la plus centrale de la rétine et est responsable de la vision centrale des couleurs à haute résolution. Les patients présentant un DR impliquant la macula ont généralement une perte visuelle plus sévère que les patients sans décollement maculaire associé. Le DR est généralement traité par la chirurgie, mais celui‐ci n'est pas toujours fructueux. Chez certains patients, l'opération est initialement une réussite, mais le DR pourrait réapparaître des mois ou des années plus tard. Les DR les plus courants, et certains DR primaires, sont associées à une croissance et à une cicatrisation de la rétine appelée vitréorétinopathie proliférative (VRP). La seule thérapie éprouvée pour le DR associé à la VRP est une autre chirurgie qui consiste à retirer les membranes de la surface de la rétine et à injecter des agents de tamponnement dans l'œil pour maintenir en place la rétine nouvellement attachée. Les principaux agents de tamponnement disponibles aujourd'hui sont des gaz divers et des huiles de silicone. On ignore si ces agents de tamponnement sont efficaces et sûrs.

Quels sont les principaux résultats de la revue ?
Nous avons trouvé quatre essais contrôlés randomisés avec un total de 601 participants qui comparaient divers agents de tamponnement. Tous les participants ont subi une intervention chirurgicale pour traiter le DR associée à la VRP.

Il ne semble pas y avoir de différences majeures entre le C₃F₈ (un type de gaz) et l'huile de silicone en termes de netteté de la vision (acuité visuelle) ou de fixation de la rétine à la macula, la zone de forme ovale située près du centre de la rétine. L'huile de silicone pourrait être meilleur que le SF₆ (un autre type de gaz) pour la fixation de la rétine à la macula et d'autres critères de jugement à court terme.

Cette revue est‐elle à jour ?
Les chercheurs de Cochrane ont recherché des études qui avaient été publiées jusqu'au 2 janvier 2019.

Authors' conclusions

Implications for practice

Based on results from the Silicone Study, participants with retinal detachment (RD) associated with proliferative vitreoretinopathy (PVR) had good results with pars plana vitrectomy (PPV) with either C₃F₈ gas or silicone oil tamponades. There is a suggestion that C₃F₈ may have certain advantages with respect to long‐term anatomic outcomes in some participants, although the visual results appear similar between the tamponade agents. The choice of tamponade agent is usually made on an individual, patient‐by‐patient basis. Factors to be considered include the configuration of the detachment, the location of the retinal breaks, the lens status, the visual status of the fellow eye, the patient's ability to comply with postoperative positioning requirements, the patient's need to travel by air in the early postoperative period, and individual physician and patient preferences.

As tamponade agents, C₃F₈ and silicone oil appear to have visual and anatomic advantages over SF₆, especially within the first year after surgery, but SF₆ may be a reasonable choice in certain clinical situations.

Based on the results from the Zafar 2016 study, the 1000‐centistoke silicone oil compared with 5000‐centistoke silicone oils or the heavy silicone oil mixture used in the HSO Study (a mixture of perfluorohexyloctane (F₆H₈) and silicone oil) does not offer any additional benefits relative to standard silicone oil (either 1000 centistokes or 5000 centistokes, per the surgeon's preference) in participants with complex or recurrent RD associated with PVR.

Implications for research

The Silicone Study delineated various relative advantages and disadvantages of 1000‐centistoke silicone oil, SF₆, and C₃F₈ as tamponade agents. The study that evaluated 1000‐centistokesilicone oil versus of 5000‐centistoke silicone oil, had high overall risk of bias and found no difference between the two groups for most of the outcomes assessed, thus a prospective clinical trial evaluated these comparisons appears warranted. Future research may develop alternative tamponade agents, particularly with a density greater than water, which would reduce the postoperative positioning requirements for many patients. Properties of an ideal tamponade agent include optical clarity, lack of toxicity, no effect on the eye's refractive state, no effect on intraocular pressure (IOP) or cataract formation, inhibition of cellular migration, and inhibition of gliosis or glial proliferation.

Summary of findings

Open in table viewer

Summary of findings 1. Silicone oil compared to sulfur hexafluoride (SF6) for surgery for retinal detachment associated with proliferative vitreoretinopathy

Outcomes	Risk with sulfur hexafluoride (SF6)	Risk with Silicone oil	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comment
Silicone oil compared to sulfur hexafluoride (SF6) for surgery for retinal detachment associated with proliferative vitreoretinopathy
Patient or population: surgery for retinal detachment associated with proliferative vitreoretinopathy Setting: eye hospital Intervention: silicone oil Comparison: sulfur hexafluoride (SF6)
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comment
Visual acuity ≥ 5/200 at two years	325 per 1,000	510 per 1,000 (302 to 865)	RR 1.57 (0.93 to 2.66)	87 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
Macular attachment at two years	575 per 1,000	788 per 1,000 (581 to 1,000)	RR 1.37 (1.01 to 1.86)	87 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
Retina detachment at two years	See comment	‐	‐	‐	‐	This outcome was not reported.
Visual acuity worse than 20/200 (regardless of anatomic outcome) at two years	See comment	‐	‐	‐	‐	This outcome was not reported.
Intraocular pressure greater than 21 mmHg at two years	See comment	‐	‐	‐	‐	This outcome was not reported.
Visually significant cataract at two years	See comment	‐	‐	‐	‐	This outcome was not reported.
Quality of life measures at two years evaluated using validated scale as reported by study	See comment	‐	‐	‐	‐	This outcome was not reported.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.
¹ Downgraded one level for risk of bias ² Downgraded one level for imprecision (as based on 1 RCT with n = 87)

Open in table viewer

Summary of findings 2. Silicone oil compared to perfluropropane (C3F8) for surgery for retinal detachment associated with proliferative vitreoretinopathy

Outcomes	Risk with perfluropropane (C3F8)	Risk with Silicone oil	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comment
Silicone oil compared to perfluropropane (C3F8) for surgery for retinal detachment associated with proliferative vitreoretinopathy
Patient or population: surgery for retinal detachment associated with proliferative vitreoretinopathy Setting: eye hospital Intervention: silicone oil Comparison: perfluropropane (C3F8)
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comment
Visual acuity ≥ 5/200 at 3 years	406 per 1,000	394 per 1,000 (296 to 532)	RR 0.97 (0.73 to 1.31)	264 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
Macular attachment at 3 years	739 per 1,000	739 per 1,000 (636 to 850)	RR 1.00 (0.86 to 1.15)	264 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
Retina detachment at 3 years	See comment	‐	‐	‐	‐	This outcome was not reported.
Visual acuity worse than 20/200 (regardless of anatomic outcome) at two years	See comment	‐	‐	‐	‐	This outcome was not reported.
Intraocular pressure (IOP) greater than 21 mmHg	See comment	‐	‐	‐	‐	This outcome was not reported.
Visually significant cataract at 3 years	See comment	‐	‐	‐	‐	This outcome was not reported.
Quality of life measures at 3 years evaluated using validated scale as reported by study	See comment	‐	‐	‐	‐	This outcome was not reported.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.
¹ Downgraded one level for risk of bias ² Downgraded one level for imprecision

Open in table viewer

Summary of findings 3. Standard silicone oil compared to heavy silicone oil for surgery for retinal detachment associated with proliferative vitreoretinopathy

Outcomes	Risk with heavy silicone oil	Risk with Standard silicone oil	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comment
Standard silicone oil compared to heavy silicone oil for surgery for retinal detachment associated with proliferative vitreoretinopathy
Patient or population: surgery for retinal detachment associated with proliferative vitreoretinopathy Setting: eye hospital Intervention: Standard silicone oil Comparison: heavy silicone oil
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comment
Change in visual acuity (logMAR) at one year	The mean change in acuity in the heavy oil group was 1.24 logMAR	MD ‐0.03 lower (‐0.35 lower to 0.29 higher)	‐	93 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}	Change in visual acuity as a dichotomous outcome was not reported, instead, the investigators reported mean change in visual acuity and rates of recurrent RD.
Macular attachment at one year	See comment	‐	‐	‐	‐	This outcome was not reported.
Retinal detachment at one year	250 per 1,000	223 per 1,000 (135 to 370)	RR 0.89 (0.54 to 1.48)	186 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
Visual acuity worse than 20/200 (regardless of anatomic outcome) at one year	See comment	‐	‐	‐	‐	Visual acuity was not reported as a dichotomous outcome.
Intraocular pressure (IOP) greater than 21 mmHg at one year	See comment	‐	‐	‐	‐	This outcome was not reported.
Visually significant cataract at one year	See comment	‐	‐	‐	‐	This outcome was not reported.
Quality of life measures at two years evaluated using validated scale as reported by study at one year	See comment	‐	‐	‐	‐	This outcome was not reported.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; MD: Mean difference
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.
¹ Downgraded one level for risk of bias ² Downgraded one level for imprecision (as based on 1 RCT with n = 93)

Open in table viewer

Summary of findings 4. 5000‐centistoke compared to 1000‐centistoke for surgery for retinal detachment associated with proliferative vitreoretinopathy

Outcomes	Risk with 1000‐centistoke	Risk with 5000‐centistoke	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comment
5000‐centistoke compared to 1000‐centistoke for surgery for retinal detachment associated with proliferative vitreoretinopathy
Patient or population: surgery for retinal detachment associated with proliferative vitreoretinopathy Setting: eye hospital Intervention: 5000‐centistoke Comparison: 1000‐centistoke
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comment
Change in visual acuity (logMAR) at 18 months	BCVA improved or remained unchanged in 77 participants (90.6%) No results were presented per intervention group, but authors reported that there was no statistically significant difference between intervention groups.			85 (1 RCT)	⊕⊕⊝⊝ LOW 1 2
Macular attachment at 18 months	Reattachment of retina was reported as successful in 67 participants (78.8%) with first surgery, and 79 participants (92.9%) with the second surgery. Authors reported no between‐group difference was observed in this outcome			85 (1 RCT)	⊕⊕⊝⊝ LOW 1 2
Retina detachment ‐ After removal of silicone oil at 18 months	136 per 1,000	49 per 1,000 (11 to 228)	RR 0.36 (0.08 to 1.67)	85 (1 RCT)	⊕⊕⊝⊝ LOW 1 2
Visual acuity worse than 20/200 (regardless of anatomic outcome) at 18 months	See comment	‐	‐	‐	‐	Visual acuity was not reported as a dichotomous outcome.
Intraocular pressure (IOP) greater than 21 mmHg at 18 months	244 per 1,000	220 per 1,000 (100 to 474)	RR 0.90 (0.41 to 1.94)	86 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}	Elevated IOP (greater than 22 mmHg) at 18 months.
Visually significant cataract at 18 months	489 per 1,000	636 per 1,000 (435 to 924)	RR 1.30 (0.89 to 1.89)	86 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
Quality of life measures at 18 months evaluated using validated scale as reported by study	See comment	‐	‐	‐	‐	This outcome was not reported.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). BCVA: Best corrected visual acuity; CI: Confidence interval; RR: Risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.
¹ Downgraded one level for risk of bias ² Downgraded one level for imprecision (as based on 1 RCT with n = 85)

Background

Description of the condition

Introduction

Retinal detachment (RD) remains a significant cause of vision loss. A variety of surgical techniques are available to treat RD. For primary RD, these procedures have a very high rate of successful anatomic retinal reattachment (overall above 90%) (Schwartz 2004). The Scleral Buckling versus Primary Vitrectomy in Rhegmatogenous Retinal Detachment (SPR) study, which excluded many relatively straightforward cases, reported single operation success rates between 60% to 80%, depending on the subgroup, and 73% overall (Heimann 2007). Most recurrent RDs, and some primary RDs, are associated with varying degrees of proliferative vitreoretinopathy (PVR), or the growth of fibrous membranes (similar to scar tissue) along the surface of the retina, which leads to traction on the retina (TRSTC 1983).

Epidemiology

Recurrent RD with PVR occurs in about 5% to 10% of patients (Charteris 2002). Major risk factors for recurrent RD with PVR include RD in the inferior (lower) portion of the eye (Singh 1986), severe ocular trauma (Kruger 2002), and giant retinal tears (Scott 2002). Other reported risk factors for recurrent RD with PVR include the inability to identify a retinal break, the use of pars plana vitrectomy (PPV) in the initial repair, preoperative PVR, preoperative choroidal detachment, and a relatively greater use of cryopexy (Cowley 1989). Recurrent RD with PVR may require multiple additional surgeries and is associated with poorer visual outcomes. These additional surgeries are associated with significantly increased costs (Patel 2004). Some patients with primary RD may also present with PVR; risk factors include large or giant retinal tears, longstanding RD, and other factors (Garweq 2013).

Presentation and diagnosis

PVR is usually diagnosed within the first few months after RD surgery. Symptoms include decreased vision in the affected eye. The diagnosis is made by dilated fundus examination in the doctor's office or outpatient clinic.

Description of the intervention

Vitreoretinal surgery is standard treatment for RD with PVR. Pars plana vitrectomy (PPV), removal of the epiretinal membranes; treatment of the retinal breaks; and injection of a tamponade agent are performed. In some cases, removal of the lens (either the crystalline lens or a previously placed intraocular lens) is performed. Tamponade is necessary to reduce the rate of fluid flow through open retinal tears, which would cause recurrent RD. The major tamponade agents available today are various gases and silicone oils. Currently available gases include air, sulfur hexafluoride (SF₆), hexafluroethane (C₂F₆), and perfluropropane (C₃F₈). The major advantage of gas tamponade is that the gas spontaneously dissipates, usually over several weeks. Currently available silicone oils come in 1000‐centistoke and 5000‐centistoke viscosities. Silicone oil is permanent and may eventually require surgical removal.

There are several investigational tamponade agents, including polydimethylsiloxane (PDMS) 1000 (Tognetto 2005), perfluorohexylethan (O62) (Hoerauf 2005), perfluoro‐n‐octane (PFO) (Rofail 2005), a mixture of perfluorohexyloctane (F₆H₈) in silicone oil (Stappler 2008), and a mixture of perfluorohexyloctane (F₆H₈) in PDMS 1000 (Heimann 2008; Tognetto 2008). Various tamponade agents with a specific gravity greater than that of water have shown evidence of toxicity in animal models, in rat retinal cell cultures in vitro, and in clinical reports (Eckardt 1990; Matteucci 2007; Singh 2001). These investigational agents are not available for routine clinical use in the USA.

Tamponade agents are useful in four broad categories of patients with RD.

Patients with primary RD, treated with PPV as a first‐line procedure. These patients are usually treated with gas tamponade rather than silicone oil.
Patients with complex or recurrent RD associated with PVR. These patients are the focus of this review. These patients are typically treated with either gas or silicone oil.
Patients with RD associated with a giant retinal tear. These patients are treated with either gas or silicone oil.
Patients with inferior RD, treated with PPV as a first‐line procedure. Some surgeons use heavy liquids, such as PFO or heavy silicone oil, as investigational agents in these patients.

How the intervention might work

Tamponade agents are believed to work by reducing or eliminating fluid vectors through open retinal breaks until the applied retinopexy (typically photocoagulation or cryopexy) creates a permanent seal. Gases such as SF₆ and C₃F₈ spontaneously dissipate, while silicone oil is permanent and may eventually require removal.

Why it is important to do this review

The various tamponade agents offer different advantages and disadvantages in terms of safety and effectiveness (Krzystolik 2000; Young 2005). It is over five years since the last version of this systematic was published (Schwartz 2014), hence an update was needed to evaluate both earlier and more recent evidence on the relative safety and effectiveness of various tamponade agents used with surgery for retinal detachment (RD) complicated by proliferative vitreoretinopathy (PVR).

Objectives

The objective of this review was to assess the relative safety and effectiveness of various tamponade agents used with surgery for retinal detachment (RD) complicated by proliferative vitreoretinopathy (PVR).

The specific comparisons depended on the RCTs we identified in the search. The secondary objectives of the review were to examine quality of life measures such as patient satisfaction and subjective visual improvement, and to summarize economic data such as direct and indirect costs of surgery and rehabilitation. We intended to compare:

the various gas tamponade agents with each other;
the two silicone oil preparations with each other;
the various gas agents versus the various silicone oils;
the established agents (gases, silicone oil) versus the investigational agents.

Methods

Criteria for considering studies for this review

Types of studies

We included randomized controlled trials (RCTs) only. We set no limitations on the various treatment arms compared.

Types of participants

We included RCTs in which participants underwent surgical repair of RD associated with PVR. We employed no restrictions with respect to age or cause of RD.

Types of interventions

We included RCTs that studied agents used as tamponade in the treatment of RD associated with PVR, such as air, sulfur hexafluoride (SF₆), hexafluroethane (C₂F₆), perfluropropane (C₃F₈), and silicone oil, as well as investigational agents such as heavy silicone oil (polydimethylsiloxane 1000), perfluorohexylethan (O62), and perfluoro‐n‐octane (PFO).

Types of outcome measures

Primary outcomes

The primary outcome for this review was visual acuity at one year. We analyzed outcomes at additional times of follow‐up as reported in the included RCTs. We intended to compare visual acuity as a dichotomous outcome (the proportion of participants who lost three or more lines of logMAR visual acuity; participants who lost one or two lines of logMAR visual acuity were considered stabilized), and also as a continuous outcome (mean logMAR scores). We considered other dichotomous and continuous visual acuity outcomes at other time points as reported in the included RCTs.

Secondary outcomes

The secondary outcome for this review was macular attachment at one year. This was chosen because in some patients with PVR complete retinal re‐attachment is not possible, but macular attachment yields generally better visual results than does persistent macular detachment. We also presented secondary outcomes measured at other time points as reported in the included RCTs.

Adverse effects (severe and minor)

Severe

Retina detached at one year
Visual acuity worse than 20/200 (regardless of anatomic outcome)

Minor

Intraocular pressure (IOP) greater than 21 mmHg
Visually significant cataract

Quality of life measures

We intended to examine patient satisfaction, subjective visual improvement, and other quality of life measures evaluated using a validated scale.

Economic data

We intended to summarize direct and indirect costs of surgery and rehabilitation and any other economic data in the included studies.

Follow‐up

We restricted studies to those with at least one year of follow‐up. We believe that shorter follow‐up periods are less clinically relevant.

Search methods for identification of studies

Electronic searches

We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (which contains the Cochrane Eyes and Vision Trials Register) 2019, Issue 1, part of the Cochrane Library (www.thecochranelibrary.com) (searched 2 January 2019), Ovid MEDLINE, Ovid MEDLINE In‐Process and Other Non‐Indexed Citations, Ovid MEDLINE Daily, Ovid OLDMEDLINE (January 1946 to January 2019June 2013), Embase (January 1980 to January 2019), Latin American and Caribbean Health Sciences Literature Database (LILACS) (January 1982 to January 2019), the metaRegister of Controlled Trials (mRCT) (www.controlled‐trials.com), ClinicalTrials.gov (www.clinicaltrials.gov) and the WHO International Clinical Trials Registry Platform (ICTRP) (www.who.int/ictrp/search/en). We did not use any date or language restrictions in the electronic searches for trials. We last searched the electronic databases on 2 January 2019.

See: Appendices for details of search strategies for CENTRAL (Appendix 1), MEDLINE (Appendix 2), Embase (Appendix 3), LILACS (Appendix 4), mRCT (Appendix 5), ClinicalTrials.gov (Appendix 6) and the ICTRP (Appendix 7).

Searching other resources

We searched the reference lists of the studies included in the review for other potential inclusions. We did not search conference proceedings for the purpose of this review. Although we initially did not intend to contact individuals or organizations specifically for this review, because we did not believe that doing so would add significantly to the data obtainable through published trials, we contacted the investigators of included studies for clarification of methods and other data reported in published manuscripts.

Data collection and analysis

Selection of studies

At least two review authors, working independently, reviewed the titles and abstracts resulting from the searches. Two review authors reviewed the full‐text manuscripts of all possibly or definitely relevant studies to determine eligibility for inclusion. We resolved any discrepancies through discussion when screening titles and abstracts and assessing the eligibility for full‐text reports. We did not mask RCT details in this process. For any unclear information, we contacted the study investigators for further clarification. We recorded the studies that we excluded during full‐text assessment, and described the reasons for exclusion in the 'Characteristics of excluded studies' table.

Data extraction and management

Extraction of study characteristics

We extracted the following information for each RCT.

Methods: method of allocation, masking (blinding), exclusions after randomization, losses to follow‐up and compliance, unusual study design.
Participants: country where participants enrolled, number randomized, age, sex, inclusion and exclusion criteria.
Interventions: test intervention, comparison intervention (control), duration of intervention.
Outcomes: visual acuity, macular attachment, complication rates, adverse effects, quality of life, and economic outcomes.
Notes: additional details (such as funding sources).

Data extraction and entry

Two review authors, working independently, extracted data using a paper data extraction form developed and piloted by Cochrane Eyes and Vision. We resolved discrepancies by discussion. One review author entered the data into RevMan 5.3 (RevMan 2014), and a second review author verified the data entry. The main outcome measures were visual acuity, macular attachment, and various complication rates. This included dichotomous data (such as retinal detachment, proportion of participants who lost three or more lines of logMAR visual acuity), as well as continuous data (such as mean logMAR visual acuity).

Assessment of risk of bias in included studies

We reviewed the risk of bias of included studies as outlined in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). At least two review authors assessed the risk of bias for each included study according to the following criteria.

Selection bias (randomized sequence generation and allocation concealment).
Performance bias (masking of participants and researchers).
Attrition bias (incomplete outcome data adequately addressed).
Detection bias (masking of outcome assessors).
Reporting bias (free of selective outcome reporting).

We judged each area of potential bias as low risk of bias, high risk of bias, or unclear risk of bias. We considered methods such as central randomization and use of sequential opaque envelopes as evidence of adequate allocation concealment. We evaluated any exclusions after randomization, losses to follow‐up and differential reasons for losses to follow‐up in the treatment groups. Any discrepancies were resolved through discussion.

We recognized that masking of participants and surgeons (performance bias) and masking of persons assessing retinal detachment (detection bias) may not be possible in studies comparing gas to silicone oil. However, studies that had successfully masked outcome data (such as studies in which visual acuity was measured by an examiner masked to the tamponade agent) were emphasized.

Measures of treatment effect

We reported unpooled risk ratios (RRs) with 95% confidence intervals (CIs) for the dichotomous outcomes of visual acuity and macular attachment for Silicone Study 1992a and Silicone Study 1992b; RR with 95% CI for recurrent RDs and mean difference (MD) with 95% CI for visual acuity for HSO Study, and RR with 95% CI for RDs (after first and second surgery, and after the removal of silicone oil), as well as elevated IOP for Zafar 2016. If other continuous outcomes are included in future updates of the review, we will calculate MDs or standardized mean differences (SMDs) depending on the types of measurement scales used.

We initially intended to compare 'all gases' (that is, SF₆, C₂F₆, and C₃F₈) versus silicone oil, but the included studies did not compare tamponade agents in this manner. Specifically, the Silicone Study conducted two RCTs, one comparing silicone oil (1000 centistokes) with SF₆, and one comparing silicone oil (1000 centistokes) with C₃F₈, another study compared standard silicone oil (either 1000 centistokes or 5000 centistokes) with heavy silicone oil (a mixture F₆H₈ and silicone oil), and the fourth compared different viscosities of silicone oil (1000 centistokes or 5000 centistokes). Accordingly, this review used the same comparisons.

Unit of analysis issues

The unit of analysis for outcomes was eyes of individuals. All four RCTs included only one eye per participant. For future updates of the review, if a RCT randomized one eye to one tamponade group and the other eye of the same person to the other group, we will only include such a design when it appropriately considered intra‐person correlation in their analyses. We will refer to the guidelines in Chapter 9 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011).

Dealing with missing data

We contacted the primary authors of the included studies to provide 12‐month visual acuity and macula status outcome data when not reported in the published papers. We did not impute data for this review, but we will consider imputation for future updates of the review and discuss the assumptions made during imputation.

Assessment of heterogeneity

We intended to assess for statistical heterogeneity using the Chi² test and the I² statistic, but since no pooled estimates were included, these assessments of heterogeneity were not applicable. If data synthesis is considered at the time of an update to this review, we will follow the following guidelines. We will consider an I² value greater than 50% to indicate substantial statistical heterogeneity. In such a situation we will not report a pooled estimate. We also will not report a pooled estimate when clinical or methodological heterogeneity (from details listed in the Characteristics of included studies table) is detected. Instead, we will report a narrative or tabulated summary of the included studies. We will use a random‐effects model to incorporate the heterogeneity if the I² value is less than 50%, unless there are fewer than three studies. If we detect no statistical heterogeneity (I² value of 0), or there are fewer than three studies, we will use a fixed‐effect model.

Assessment of reporting biases

We assessed selective outcome reporting by comparing outcomes listed in the protocol of the RCTs and the outcomes analyzed in the final published report. For future updates of the review, when the protocol of an included study is not available, we will compare the outcomes pre‐specified in the methods section and outcomes analyzed in the results section, and will follow the guidelines in Chapter 10 of the Cochrane Handbook for Systematic Review of Interventions (Sterne 2011). We plan to examine funnel plots from each meta‐analysis to assess reporting bias when at least 10 studies are included.

Data synthesis

No pooled estimates of included studies are reported. If pooled estimates are considered for future updates of the review, we will follow the guidelines in Chapter 9 of the Cochrane Handbook for Systematic Review of Interventions (Deeks 2011).

Subgroup analysis and investigation of heterogeneity

We will consider subgroup analyses, as appropriate, in future updates of this review, and will consult the guidelines for investigating heterogeneity in Chapter 9 of the Cochrane Handbook for Systematic Review of Interventions (Deeks 2011). One possible strategy is to divide participants by surgical history, such as participants with chronic RD with PVR and no previous surgery, participants with recurrent RD following scleral buckling only, and participants with recurrent RD following PPV and previous intravitreal tamponade (gas or oil). Another possible strategy is to divide participants with certain high‐risk clinical features, such as participants with giant retinal tear, participants with open‐globe trauma, and participants under 18 years of age.

Sensitivity analysis

We planned to examine the impact of the exclusion of unpublished and industry‐funded studies in sensitivity analyses, but these exclusions were not applicable to the current systematic review.

'Summary of findings' tables

We presented a 'Summary of findings' table for each comparison of interest when data were available, including strengths and limitations of evidence for primary, secondary, and adverse outcomes. Two review authors independently graded the overall certainty of the evidence for each outcome using the GRADE classification (www.gradeworkinggroup.org). We assessed the certainty of evidence for each outcome as 'high,' 'moderate,' 'low,' or 'very low' according to the following criteria as described in Chapters 11 and 12 of the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2011a; Schünemann 2011b).

The following comparisons were included: 1) silicone oil (1000 centistokes) versus sulfur hexafluoride (SF₆); 2) silicone oil (1000 centistokes) versus perfluropropane (C₃F₈); 3) Standard silicone oil (either 1000 centistokes or 5000 centistokes) with heavy silicone oil (a mixture F₆H₈ and silicone oil); 4) 5000 centistokes versus 1000 centistokes. For each comparison, the following outcomes at follow‐up time point ≥1 year as defined by each study post‐treatment are included in the 'Summary of findings' tables.

Visual acuity ≥ 5/200
Macular attachment
Retina detached
Visual acuity worse than 20/200 (regardless of anatomic outcome) at two years
Intraocular pressure (IOP) greater than 21 mmHg
Visually significant cataract
Quality of life measures at two years evaluated using validated scale as reported by study

Results

Description of studies

Results of the search

Detailed results of the previous search were published in the 2014 version of this review (Schwartz 2014). Briefly, three RCTs were included in total and one record was classified as awaiting classification after screening 108 records from updated search on 26 June 2013. In January 2019, an updated electronic literature search yielded 934 additional records. After duplicate removal, 929 titles and abstracts were screened by two review authors independently, of which 24 relevant full‐text reports were identified. Of these, 17 full‐text reports were excluded with reasons. Five studies were classified as ongoing, one was classified as awaiting classification and one new RCT (Zafar 2016) added in this update (Figure 1).

Figure 1

Study flow diagram.

Overall, we included four RCTs, excluded 37 studies (37 records), classified two studies (two records) as awaiting classification and five assessed as ongoing or completed with results not yet published.

Included studies

We identified four RCTs that met our inclusion criteria. Two RCTs (from the Silicone Study) were conducted in the USA. Enrollment for the first RCT comparing silicone oil to SF₆ gas occurred from September 1985 to September 1987 (Silicone Study 1992a). For the second part of the study period, SF₆ gas was replaced with the longer‐lasting C₃F₈ gas. Enrollment for the second RCT comparing silicone oil to C₃F₈ occurred between September 1987 to October 1990 (Silicone Study 1992b). Participants aged 18 years or older and with retinal detachment (RD) associated with proliferative vitreoretinopathy (PVR) were offered randomization. One eye per patient was randomized and grouped as eyes that had not undergone prior vitrectomy (Group 1) or eyes that had undergone vitrectomy but without silicone oil injection (Group 2). The first RCT included 113 eyes in Group 1 and 38 eyes in Group 2; the second RCT included 132 eyes in Group 1 and 139 eyes in Group 2. The exclusion criteria were uncontrolled concomitant eye disease, a history of blunt trauma within three months of entry into the study, a history of penetrating trauma, a giant retinal tear of 90 ° or greater, proliferative diabetic retinopathy, and any medical condition that could preclude participation in a three‐year study.

The Heavy Silicone Oil Study (HSO Study), compared vitrectomy with heavy silicone oil (a mixture of perfluorohexyloctane (F₆H₈) and silicone oil) versus standard silicone oil (either 1000 centistokes or 5000 centistokes, per the surgeon's preference) and was performed between December 2003 and February 2008. The HSO Study was a multi‐center study conducted in Germany, Austria, Sweden, the UK, China, Poland, Portugal, the Netherlands, Italy, Hungary, and the USA. Ninety‐four participants with RD associated with inferior and posterior PVR or inferior RD with inferior giant retinal tear were randomized into the two intervention groups, with 46 participants in the heavy silicone oil group and 48 in the standard silicone oil group. The exclusion criteria included: RD associated with superior anterior PVR; superior giant retinal tear; retinotomies; holes or tears between 10 and 2 o’clock; diabetic retinopathy requiring treatment; glaucoma resulting in visual field defects requiring treatment; no written informed consent; age below 18 years; participation in another clinical trial; or pregnancy.

The newly included study (Zafar 2016) compared 1000‐centistoke silicone oil versus 5000‐centistoke silicone oil among 85 patients with superior rhegmatogenous retinal detachments associated with PVR grades B and C, which involves not more than 3 clock hours. It was conducted from January 2007 to June 2013 in Pakistan. Patients with history of any intra‐ocular surgery, pre‐existing glaucoma, inflammatory eye condition, traumatic RD, intra‐ocular foreign bodies, aphakia and with any pre‐existing retinopathy, with eyes in which the retina could not be re‐attached at the time of surgery, were excluded. Patients with less than 18 months of follow‐up or had incomplete records were excluded from the analysis.

The Silicone Study was funded by the National Eye Institute, National Institutes of Heath, USA, and the HSO Study was funded by the German Research Foundation (Deutsche Forschungsgemeinschaft). Zafar 2016 did not report sources of funding. None of the study investigators reported declaration of interests.

Excluded studies

We excluded 37 records altogether and listed them in the 'Characteristics of excluded studies' table with reasons for exclusion. Twenty‐four of the 37 records were not RCTs, five of them were conducted in a population that is not of interest to this review, four did not use an intervention of interest, three had a follow‐up duration less than one year and the remaining study was a conference abstract classified in previous review as awaiting classification, and a full report never got published.

Ongoing studies and studies awaiting classification

We classified two studies as awaiting classification owing to insufficient information to determine eligibility. One was an American Academy of Ophthalmology (AAO) abstract (Oncel 2006) with no published full text, and the other study did not provide enough information to permit judgement for eligibility (Trepsat 1987). They are listed in the 'Characteristics of studies awaiting classification' table. We identified five studies that were either ongoing studies (NCT02988583) or completed more than six years ago with results not yet published (see Characteristics of ongoing studies).

Risk of bias in included studies

Four RCTs met the inclusion criteria for this review (HSO Study; Silicone Study 1992a; Silicone Study 1992b; Zafar 2016). Since the two RCTs of the Silicone Study were part of the same study protocol, they followed the same design, methods, and analyses (Azen 1991 in Silicone Study 1992a). Both the Silicone Study and HSO Study were of good methodological quality and at low risk of bias (Figure 2) except that whether the participants were masked was not reported explicitly in any of these RCTs, and the Silicone Study did not mask all outcome assessors. Zafar 2016 did not provide sufficient information to judge risk of bias in most domains.

Figure 2

Methodological quality summary: review authors' judgements about each methodological quality item for each included study.

Allocation

For the Silicone Study, the randomization scheme was administered centrally through the Data Coordinating Center and employed stratification and blocking to ensure equal treatment assignments within each clinical center. Treatment allocation was adequately concealed with sequential opaque envelopes delivered to each study site and opened at the time of tamponade injection. For the HSO Study, randomization was generated using permuted blocks of varying sizes, stratified by surgeon. Treatment allocation was adequately concealed with sealed envelopes opened after study enrollment. Method of random sequence generation and allocation concealment was not reported in Zafar 2016.

Masking (performance bias and detection bias)

None of the RCTs reported masking of participants or surgeons. The study outcome assessors and surgeons were not masked for the two RCTs of the Silicone Study, but were masked in the HSO Study. Masking was not reported in one study (Zafar 2016).

Incomplete outcome data

For the Silicone Study, the last observation carried forward method was used for missing data. Data were imputed for participants who missed intermediate examinations, but attended prior and subsequent examinations, only when findings were deemed consistent. In the event that a retinal detachment recurred during the study period and required surgery, participants were analyzed using the original random treatment allocation. Randomized participants from a study centre that ceased follow‐up during the study period were excluded from the analysis (12 out of 151 participants from Silicone Study 1992a and six out of 271 participants from Silicone Study 1992b). However, the first RCT (Silicone Study 1992a) also excluded 38 participants who had previous vitrectomy from the final analyses, therefore, almost a third of participants in this study did not contribute to any outcome data (51 participants out of 151), so we assessed the risk of bias as high.

For the HSO Study, participants who did not satisfy the major inclusion criteria but were already randomized (performed preoperatively) were all included in the full analysis set; three participants in the heavy silicone oil group and five participants in the standard silicone group fulfilled intraoperative exclusion criteria. One participant was excluded from analysis due to a lack of pre‐ and post‐surgical assessment and data (only randomization sheet present).

For Zafar 2016, 11 participants (11.5%) participants who were lost to follow‐up less than 18 months after surgery, or had incomplete records were not included in the final analysis. We judged high risk of bias for this domain.

Selective reporting

Both the RCTs from the Silicone Study (Silicone Study 1992a; Silicone Study 1992b) appeared to be free of selective reporting since the primary and secondary outcomes were published a priori in their respective methods paper (Azen 1991 in Silicone Study 1992a). However, in the methods paper for the HSO Study (Joussen 2007 in HSO Study), it pre‐specified to measure quality of life outcomes, but no data were reported. Therefore, we assessed the reporting bias as low for the Silicone Study and high for the HSO Study. No protocol or registration was identified for Zafar 2016, therefore we assessed the reporting bias as unclear.

Other potential sources of bias

Fourteen baseline characteristics were compared between treatment arms in the Silicone Study (age, sex, study eye, prior scleral buckle, other ocular surgery, mean duration of RD, Retina Society classification, visual acuity, refractive status, intraocular pressure (IOP), corneal status, aqueous flare, aqueous cell, and neovascularization). The Silicone Study investigators reported one statistically significant difference in baseline characteristics between the eyes of participants assigned to receive SF₆ gas and those assigned to receive silicone oil (Silicone Study 1992a).The estimated duration of RD was greater in Group 2 eyes (eyes of participants with prior vitrectomy but without silicone oil injection) randomized to SF₆ compared to Group 2 eyes randomized to silicone oil. This information was not sufficient to determine whether other potential source of bias exist.We identified no other potential bias for the remaining two studies (HSO Study; Zafar 2016).

Effects of interventions

See: Summary of findings 1 Silicone oil compared to sulfur hexafluoride (SF6) for surgery for retinal detachment associated with proliferative vitreoretinopathy; Summary of findings 2 Silicone oil compared to perfluropropane (C3F8) for surgery for retinal detachment associated with proliferative vitreoretinopathy; Summary of findings 3 Standard silicone oil compared to heavy silicone oil for surgery for retinal detachment associated with proliferative vitreoretinopathy; Summary of findings 4 5000‐centistoke compared to 1000‐centistoke for surgery for retinal detachment associated with proliferative vitreoretinopathy

Silicone oil versus gas tamponades

The Silicone Study conducted two RCTs, one comparing silicone oil (1000 centistokes) with SF₆ and one comparing silicone oil (1000 centistokes) with C₃F₈. Below we have described results for each outcome we pre‐specified in the methods section of this review. For the first RCT comparing silicone oil (1000 centistokes) with SF₆, the study investigators performed statistical analyses only on non‐vitrectomized eyes (group 1) because the sample size of eyes that had already undergone vitrectomy (group 2) was small (38 participants).

Visual acuity

We intended to compare visual acuity as a dichotomous outcome (the proportion of participants who lost three or more lines of logMAR visual acuity; participants who lost one or two lines of logMAR visual acuity were considered stabilized), and also as a continuous outcome (mean logMAR scores); however, no studies reported the proportion of participants who lost three or more lines of visual acuity, instead, participants achieving 5/200 or better visual acuity were reported for both groups. The cut‐off point of 5/200 was chosen because 5/200 is considered 'ambulatory vision' (enough vision not to bump into large objects while walking) and is used in some clinical trials with severe diseases and generally bad outcomes. The Silicone Study recorded visual acuity using the Diabetic Retinopathy Vitrectomy Study protocol and charts.

Two RCTs including 352 eyes of 352 participants contributed to this outcome at 24 months (87 eyes in Silicone Study 1992a), or at the last follow‐up evaluation (264 eyes in Silicone Study 1992b). When silicone oil was compared with SF₆, the study investigators reported that eyes that had not undergone prior vitrectomy (Group 1) and were randomized to receive silicone oil more often achieved a visual acuity of 5/200 or better at one year (P < 0.05; data on visual acuity not reported), but there was no evidence of a difference between the groups at two years (1 RCT; 87 participants; risk ratio (RR) 1.57; 95% confidence interval (CI) 0.93 to 2.66) (Figure 3). The certainty of evidence was low, there were wide confidence intervals. We downgraded for risk of bias and imprecision.

Figure 3

Forest plot of comparison: 1 Silicone oil versus SF6, outcome: 1.1 Visual acuity ≥ 5/200 and macular attachment at 24 months.

Forest plot of comparison: 1 Silicone oil versus SF₆, outcome: 1.1 Visual acuity ≥ 5/200 and macular attachment at 24 months.

When silicone oil (1000 centistokes) was compared with C₃F₈, there were no evidence of a differences between the groups with respect to visual acuity of 5/200 or better at a minimum of one year, 264 participants, (RR 0.97; 95% CI 0.73 to 1.31) (Figure 4). The certainty of evidence was low, downgrading for risk of bias and imprecision.

Figure 4

Forest plot of comparison: 2 Silicone oil versus perfluropropane (C3F8), outcome: 2.1 Visual acuity ≥ 5/200 at last follow‐up examination.

Forest plot of comparison: 2 Silicone oil versus perfluropropane (C₃F₈), outcome: 2.1 Visual acuity ≥ 5/200 at last follow‐up examination.

Macular attachment

Two RCTs including 352 eyes of 352 participants contributed to this outcome at 24 months (87 eyes in Silicone Study 1992a) or at last follow‐up evaluation (264 eyes in Silicone Study 1992b). When silicone oil was compared with SF₆, the study investigators reported that eyes that had not undergone prior vitrectomy (Group 1) and were randomized to receive silicone oil were 37% more likely to achieve macular attachment at both one year (P < 0.05; data on macular attachment not reported). At two years, participants receiving silicone oil compared with SF₆, may experience slight improvement in macular attachment, (1 RCT; 87 participants; RR 1.37; 95% CI 1.01 to 1.86) (Figure 3). The certainty of evidence was low, sample size was small. We downgraded for risk of bias and imprecision.

When silicone oil (1000 centistokes) was compared with C₃F₈, there was no evidence of a differences between the groups with respect to macular attachment at a minimum of one year follow‐up (1 RCT; 264 participants; RR 1.00; 95% CI 0.86 to 1.15) (Figure 5). However, the proportions of eyes with postoperative macular attachment were higher in eyes randomized to C₃F₈ versus silicone oil at each time point, and this difference favored the C₃F₈ group at 36 months (83% versus 60%; P = 0.045; standard deviation or 95% CI not provided). The certainty of evidence was low, downgrading for risk of bias and imprecision.

Figure 5

Forest plot of comparison: 2 Silicone oil versus perfluropropane (C3F8), outcome: 2.2 Macular attachment at last follow‐up examination.

Forest plot of comparison: 2 Silicone oil versus perfluropropane (C₃F₈), outcome: 2.2 Macular attachment at last follow‐up examination.

Adverse effects (severe, minor)

Two RCTs comprising 366 eyes of 366 participants contributed to the adverse event outcomes.

Severe (retina detached at one year and visual acuity worse than 20/200)

Retina detachment and visual acuity worse than 20/200 were not reported in the two RCTs.

Minor (intraocular pressure (IOP) greater than 21 mmHg and visually significant cataract)

Intraocular pressure greater than 21 mmHg was not reported in the RCTs, however, IOP greater than or equal to 30 mmHg was reported in one eye treated with SF₆ gas and no eyes treated with silicone oil in the first RCT, and two eyes treated with C₃F₈ gas and one eye with the silicone oil for the second RCT (RR 1.04; 95% CI 0.10 to 11.40) (Silicone Study 1992b) during the follow‐up for three years. The certainty of evidence was low, there were wide confidence intervals. We downgraded for risk of bias and imprecision.

Visually significant cataract

SF₆, C₃F₈, and silicone oil can worsen cataracts. However, it was unlikely that cataract progression played a major role in the visual outcomes because most eyes were pseudophakic or aphakic at one year. In the silicone oil versus SF₆ study, about 40% of the eyes were phakic at baseline, and the lens was subsequently removed in 69% of the eyes in the silicone oil group and 90% in the SF₆ group for the non‐vitrectomized eyes (RR 0.76; 95% CI 0.53 to 1.09) (Silicone Study 1992a). In the silicone oil versus C₃F₈ study, 48% of eyes were phakic at baseline, and the lens was subsequently removed in 91% of these eyes in the silicone oil group and 86% in the C₃F₈ gas group for the non‐vitrectomized eyes (RR 1.06; 95% CI 0.88 to 1.26), and in 93% of the eyes in the silicone oil group and 100% in the C₃F₈ group (1 RCT; 87 participants; RR 0.71; 95% CI 0.53 to 0.95) (Silicone Study 1992b). The certainty of evidence was low. We downgraded for risk of bias and imprecision.

Quality of life measures

The Silicone Study did not specifically address quality of life measurements.

Economic data

The Silicone Study did not specifically address economic analysis, but a subsequent economic model including data from the Silicone Study reported that surgery for retinal detachment (RD) associated with proliferative vitreoretinopathy (PVR) was cost‐effective. In eyes that had not undergone previous pars plana vitrectomy (PPV), silicone oil (USD per quality‐adjusted life year (QALY) gained of USD 40,252) was slightly more cost‐effective than C₃F₈ (USD per QALY gained of USD 46,926). In eyes that had undergone previous PPV, C₃F₈ (USD per QALY gained of USD 46,162) was more cost‐effective than silicone oil (USD per QALY gained of USD 62,383) (Brown 2002).

Standard silicone oil versus heavy silicone oil

The HSO Study conducted one RCT, comparing standard silicone oil (either 1000 centistokes or 5000 centistokes, per the surgeon's preference) and heavy silicone oil (a mixture of perfluorohexyloctane (F₆H₈) and silicone oil) in participants with inferior RD associated with PVR.