Article Text

Download PDFPDF

Scleral buckling versus pars plana vitrectomy in simple phakic macula-on retinal detachment: a propensity score-matched, registry-based study
  1. Sumihiro Kawano1,
  2. Takumi Imai2,
  3. Taiji Sakamoto3
  4. the Japan-Retinal Detachment Registry Group
    1. 1 Ophthalmology, Kurashiki Central Hospital, Kurashiki, Okayama, Japan
    2. 2 Medical Statistics, Osaka City University Graduate School of Medicine School of Medicine, Osaka, Japan
    3. 3 Ophthalmology, Kagoshima University, Kagoshima, Japan
    1. Correspondence to Dr Taiji Sakamoto, Ophthalmology, Kagoshima University, Kagoshima 890-0065, Japan; tsakamot{at}


    Background/Aims To assess the anatomical and visual results of uncomplicated phakic macula-on retinal detachment (RD) in patients treated with pars plana vitrectomy (PPV) or scleral buckling (SB).

    Methods A retrospective cohort of patients aged <65 years and diagnosed with uncomplicated phakic macula-on primary RD, who were registered in the Japan-Retinal Detachment Registry, was compiled between February 2016 and March 2017. We performed propensity score matching using preoperative findings and surgeon-related factors as covariates to account for relevant confounders. The primary outcome was anatomical failures at 6 months postoperatively, classified as follows: level 1, an inoperable state; level 2, anatomical recovery with silicone oil tamponade; and level 3, need for additional surgery to repair the detachments. The secondary outcome was change in best-corrected visual acuity (BCVA).

    Results Of the 822 included patients, 552 underwent PPV and 270 underwent SB. After propensity score matching, 137 matched cases between the PPV and SB groups were analysed. The total proportion of surgical failures in the PPV group was higher than that in the SB group (risk difference, 0.10 (0.02 to 0.18), p=0.011, McNemar’s test). Conversely, the change in BCVA was not significantly different between the two groups (logMAR units, −0.015 (−0.084 to 0.053), p=0.66, paired samples t-test).

    Conclusions Although the indications for PPV are becoming broader, PPV may not be the optimal approach for repairing all types of RD. Therefore, careful consideration is needed when selecting the right surgical technique for treating uncomplicated phakic macula-on RD cases.

    • retina
    • treatment surgery

    Data availability statement

    No data are available.

    Statistics from

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


    Rhegmatogenous retinal detachment (RD) is a severe ophthalmological condition, which could lead to permanent visual impairment. Treatments for RD include pars plana vitrectomy (PPV), scleral buckling (SB)1 and pneumatic retinopexy, which have been compared in numerous studies.2–8 Due to the higher proportion of anatomical success relative to that with pneumatic retinopexy,9 PPV and SB have been chosen for the majority of surgical interventions in Japan, and their surgical techniques have been advancing within recent years; for example, SB with a non-contact wide-angle viewing system10 11 and PPV with microincision vitrectomy surgery (MIVS) have been developed.

    As PPV with MIVS has contributed to a reduction in iatrogenic retinal breaks and postoperative retinal detachments,12 indication of PPV in treating RD has been increasing, and it is increasingly being used to treat patients who would have been previously treated with SB.1 13 As a result, clinical indications and surgical outcomes of the two techniques may have changed in recent years; however, such changes have been poorly studied.

    In a previous randomised clinical trial (RCT) that compared PPV and SB techniques, Heimann et al 14 included patients with RD of medium complexity in two subtrials with phakic eye and aphakic/pseudophakic eye to compare the two different techniques. In this study, patients with uncomplicated phakic macula-on RD were the target population as the probability of receiving either PPV or SB was well balanced in recent trends.

    An RCT is the gold-standard study design to establish cause-and-effect relationships based on interventions; however, RCTs are expensive to conduct and time-consuming. To overcome these limitations, in 2014, the Japan Retinal Detachment (J-RD) Registry was established and compiled between February 2016 and March 2017 for detailed information on RD. This registry is advantageous due to its sufficient preoperative and follow-up information.15–17 In addition, statistical analysis methods based on propensity score are recently available, which enable researchers to estimate the treatment outcomes by reducing or eliminating measured confounders between interventions and outcomes within observational studies.18

    In this study, we focused on uncomplicated cases of phakic RD with macula-on included in the J-RD Registry. We performed propensity score matching18 and compared the PPV and SB groups to assess the differences in the total rate of surgical failure and the best-corrected visual acuity (BCVA) at 6 months postoperatively.


    Data source

    Detailed information on the J-RD Registry data is provided elsewhere.15 Patient characteristics and preoperative characteristics were recorded as follows: age, sex, prior ocular intervention, cause of current RD, BCVA at baseline, intraocular pressure, axial length, lens status, laterality, interval from onset to surgery, macular on/off, estimated time of macular off, extent of detached retina (quadrants), choroidal detachment, number of breaks, location of the largest break, size of the largest break, types of retinal break, grade of proliferative vitreoretinopathy (PVR) and location of PVR.

    In addition, we collected data related to systemic or other eye diseases for each RD case. We recorded retinal surgeons’ information registered at the time of inception of the registry which comprised self-reported data regarding their experience in years and the number of RD cases treated.

    Ethical considerations

    The requirement for informed consent was waived because all information on individual patients is deidentified in the J-RD Registry database. Although written informed consent was obtained from all participants enrolled by Kyushu University Hospital, all information on those patients was also deidentified in the registry.

    Study design

    This was a retrospective cohort study using propensity score matching. The target population included patients with uncomplicated macula-on phakic RD.

    The detailed inclusion criteria were as follows: (1) primary RD, (2) macula-on state RD, (3) phakic state and (4) age <65 years. The exclusion criteria were as follows: (1) presence of systemic or other eye diseases that could contribute to RD or cataract or that might not allow adequate preoperative fundus assessment, such as atopic dermatitis, trauma, vitreous haemorrhage and some diseases related to hereditary retinal detachment; (2) prior ocular intervention, grouped A to D in a previous report15 (group A, surgery related to current RD; group B, vitreoretinal surgery not related to current RD; group C, intraocular surgery other than vitreoretinal disease; group D, prophylactic laser photocoagulation for RD); (3) complex RD defined by the European Vitreous and Retina Society (EVRS),19–21 comprising PVR grade B/C, and characterised by choroidal detachment, hypotony, large retinal tears (>90°) and macula hole RD; and (4) tear at the vitreous base. These patients were excluded because of the high probability of being recommended for the PPV from a medical perspective.

    After the study cohort was selected, patients with missing variables were excluded from the propensity score estimation process, which is described in the Statisitical analysis section.


    Primary endpoints included the anatomical outcome at 6 months postoperatively, assessed by the difference in the total rate of surgical failure between the PPV and SB groups. Failure was defined by the EVRS definition and classified into three categories, as follows: (1) level-1 failure, a true failure that proved to be inoperable; (2) level-2 failure, anatomical recovery with silicone oil tamponade at the study endpoint; and (3) level-3 failure, anatomical recovery at the study endpoint, but requiring additional surgery because of recurrent detachment or a complication. These were performed by each surgeon’s discretion with routine optical coherence tomography monitoring. However, we did not include cataract surgery as a complication; these cases were analysed as a secondary endpoint.

    The secondary endpoint was the difference in the change in BCVA at 6 months postoperatively between the two groups. In addition, we analysed differences in the phakic ratio immediately after the surgery and at 6 months postoperatively. Both endpoints were assessed based on the intention-to-treat principle.

    Statistical analysis

    The propensity score matching method was used to compare the outcomes between the PPV and SB groups with similar covariate distributions. We used all available RD cases in the registry. The propensity score is the probability of treatment assignment conditional on measured baseline covariates.22 This method consists of two processes: propensity score estimation and propensity score matching.

    The propensity score was estimated by a multivariable logistic regression model using patient and surgeon data as covariates18 for complete case analysis. These covariates were selected as potential confounders using directed acyclic graphs23 and putative relationships, based on previous reports.1 4 6 15 21 24–26 As patient covariates, we included age, sex, intraocular pressure, BCVA at baseline, axial length category, extent of detached retina, number of breaks as a categorical variable, type of the break, and location and size of the largest break. As surgeon covariates, we included the number of RD cases treated and the surgeon’s years of experience at baseline. For propensity score matching, the calliper of width was set to 0.2 of the SD of the logit of the propensity score.27 After matching, standardised differences were assessed to evaluate balancing of covariates between the two treatment groups.

    These covariates are presented using descriptive statistics. Continuous variables are presented as mean±SD and median (IQR). Categorical variables are presented as numbers and proportions in percentages. Anatomical outcomes were assessed using McNemar’s test in propensity score-matched cases.28 The change in BCVA was evaluated using paired samples t-tests. χ2 and Mann-Whitney U tests were performed to determine the significance of the differences, as appropriate. The threshold for statistical significance was set at p<0.05. All analyses were carried out using R V3.53 (

    Sensitivity analyses

    To confirm the robustness of the results to wider study populations, we performed sensitivity analyses. First, since a relatively high proportion of eligible patients was lost during the propensity score matching process, we performed an additional inverse probability of treatment weighting (IPTW) with stabilised weights using the propensity score22 in order to estimate the association between interventions and outcomes.

    Second, we conducted propensity score matching and IPTW analysis in the entire cohort, including patients with missing covariates, using multiple imputation methods.29


    The cohort comprised 898 patients selected from the 3178 patients with primary RD included in the J-RD Registry (figures 1 and 2). Among the 2289 excluded patients, 327 had a systemic condition or other eye disease, such as trauma (n=102), atopic dermatitis (n=83) or vitreous haemorrhage (n=32). Variables were missing for 67 patients; thus, 822 patients were included in the study population for complete case analyses in the propensity score estimation process. (table 1)

    Figure 1

    Flow chart depicting the study population. EVRS, European Vitreous and Retina Society; PPV, pars plana vitrectomy; RD, retinal detachment; SB, scleral buckling.

    Figure 2

    Distribution of the propensity scores with propensity score matching in the study population. PPV, pars plana vitrectomy; SB, scleral buckling.

    Table 1

    Baseline characteristics of the study population

    Overall, 152 patients treated with SB were matched with 152 patients treated with PPV. Finally, 137 matched pairs were analysed because 15 patients were lost for follow-up. Patient characteristics are shown in table 2. All differences in characteristics were reduced in terms of standardised differences (the highest difference was 0.18 in the matched cohort). Moreover, a non-contact wide-angle viewing system was used for 14 patients treated with SB (14/137, 10.2%). A 25-gauge port was employed for 130 patients treated with PPV (130/137, 94.9%). There was no significant difference in the operation time between the PPV and SB groups (PPV vs SB: 71 min (IQR 59 to 96) vs 72 min (IQR 52 to 109), p=0.81, Mann-Whitney U test).

    Table 2

    Baseline characteristics of the study population after propensity score matching

    Regarding the anatomical outcome (table 3), the total rate of failure was 15.3% (21/137) in the PPV group and 5.1% (7/137) in the SB group at 6 months postoperatively (risk difference, 0.10 (0.02 to 0.18), p=0.011, McNemar’s test). The PPV group included 21 level-3 failure cases, indicating that 21 patients required reoperation. The reasons for reoperation were as follows: recurrent RD (n=13), macular pucker (n=5), PVR (n=1), macular hole (n=1) and unrecorded case (n=1). Conversely, the SB group included seven level-3 failure cases. The reasons for reoperation were recurrent RD (n=6) and silicone oil removal (n=1). In one case of level-3 failure in the SB group, the patient experienced an operative complication—tobramycin migrated into the subretinal space during the procedure of sub-Tenon injection. The surgery was thus converted to an SB surgery, with PPV followed by silicone oil tamponade. We included this patient in the SB group based on the intention-to-treat principle.

    Table 3

    Proportion of failed vitrectomies vs scleral buckle in the propensity score-matched groups

    Regarding the secondary endpoint, the change in BCVA was not statistically significant between the two groups (logMAR units, PPV vs SB: −0.02 (−0.08 to 0.05), p=0.66, paired samples t-test). In the PPV group, 110 patients (110/137, 80.3%) received intraocular lens (IOL) implantation intraoperatively, while all patients in the SB group (n=137) remained in a phakic state (p<0.001, χ2 test). At 6 months postoperatively, 111 patients (111/137, 81.0%) in the PPV group and 3 (3/137, 1.20%) in the SB group had pseudophakia (p<0.001, χ2 test).

    Sensitivity analyses

    In IPTW with stabilised weight analysis (table 4), the difference in the total rate of failure between the two groups did not reach statistical significance (risk difference PPV vs SB: 0.02 (–0.03 to 0.06), p=0.46), while the change in BCVA was also not statistically significant (logMAR units, PPV vs SB: 0.03 (0.02 to 0.07), p=0.25). Estimated weights were 0.33 at minimum and 17.4 at maximum using the stabilised weight method (figure 3).

    Table 4

    Baseline characteristics of the study population after the inverse probability of treatment weighting analysis

    Figure 3

    Distribution of the propensity scores with inverse probability of treatment weighting (IPTW) analysis in the study population. Estimated weights were 0.33 at minimum and 17.4 at maximum using the stabilised weight method. PPV, pars plana vitrectomy; SB, scleral buckle.

    The second sensitivity analysis which included patients with missing covariates (67 cases in total, 67/889, 7.5%) did not show meaningful differences in the main outcomes (propensity score matching analysis PPV vs SB: risk difference in the total rate of failure: 0.08 (0.01 to 0.15), p=0.034; BCVA change logMAR units: −0.01 (−0.08 to 0.06), p=0.84; IPTW analysis PPV vs SB: risk difference in the total rate of failure: 0.01 (−0.04 to 0.05), p=0.70; BCVA change logMAR units: 0.02 (−0.02 to 0.07), p=0.30) (figure 4).

    Figure 4

    Risk difference of total surgical failures for pars plana vitrectomy and scleral buckling. IPTW, inverse probability of treatment weighting; PPV, pars plana vitrectomy; PS, propensity score; SB, scleral buckle.


    Most previous RCTs comparing PPV and SB in the treatment of RD were performed during the 1990s to the early 2000s, when PPV was performed using a 20-gauge system, and meta-analysis of phakic RD results showed that final reattachments did not differ, while better visual acuity and lower incidence of cataract were observed in the SB groups than in the PPV groups.8 In this study, since the widespread use of MIVS has changed surgical choice and outcomes of PPV and SB, we focused on patients with uncomplicated phakic macula-on RD as the target population and performed propensity score matching to assess the anatomical outcome and BCVA change at 6 months postoperatively between the PPV and SB groups.

    Regarding the anatomical outcome, reoperation was more frequent in the PPV group than in the SB group. Because of differences in the anatomical approach for treating the detached retina, patients in the PPV group experienced complications resulting from radical changes in the intraocular environment, such as macula pucker, macula hole and PVR. To our surprise, complications of PPV were relatively more frequent than those previously reported.24 30 This may be explained by differences in the target populations between our study and previous studies, as incomplete peripheral vitreous body removal due to the small extent of the detached retina in macula-on RD cases may have resulted in more complications.

    Although most PPV cases were accompanied by phacoemulsification (110/137, 80.3%), a previous report that included a large sample size did not suggest increased complications in RD cases treated by combined phaco-vitrectomy.31 Thus, we believe that phacoemulsification during the RD surgery did not contribute to RD recurrence in our study. In contrast, SB resulted in fewer complications than those previously reported,24 30 which may also be explained by differences in the target populations. As Thelen et al 32 have reported, greater success is noted for macula-on than for macula-off cases.

    Regarding the visual outcome, there was no significant difference in the change in BCVA at 6 months postoperatively between the PPV and SB groups, and most patients in the PPV group (110/137, 80.3%) received IOL implantation with combined phaco-vitrectomy during the surgery. Although combined phaco-vitrectomy is not widely accepted in some countries,33 Shu et al 24 explained that most Japanese retinal surgeons prefer performing PPV combined with phacoemulsification in patients older than 50 years since previous reports revealed cataract progression after PPV in this patient population.34 Moreover, age-related decline of the accommodative ability of the crystalline lens35 may allow IOL implantation. In this study, combined phaco-vitrectomy may have prevented cataract progression from worsening the postoperative BCVA, by replacing phakia with IOL. As a result, the improvement in BCVA did not reach statistical significance between the two groups. Relative to that of macula-off RD cases, the postoperative BCVA of macula-on RD cases may be influenced to a greater extent by lens opacity because of the preserved macula function in the former.

    The higher complication rate in the PPV group may also be related to factors associated with surgeons. A previous RCT only included surgeons who had experience with more than 100 PPV and 100 SB surgeries6; however, we did not apply a similar rule and only balanced the surgeon covariates between the two groups. This may suggest that the PPV technique for phakic macula-on RD requires greater surgical skills and experience than does the SB technique. Surgeon-related factors in PPV may require further investigation.

    Interpretation of the sensitivity analysis

    Regarding the anatomical outcomes, both propensity score matching and IPTW with stabilised weight analysis showed fewer failure outcomes in the SB group than in the PPV group. However, the result using the IPTW method was statistically insignificant, denoting that the results are inconclusive (figure 4).

    Propensity score matching was used to compare the outcomes between the PPV and SB groups with similar covariate distributions, while IPTW was used to compare the outcomes in a pseudo-population where every patient within the population was offered either PPV or SB. Different statistical results between IPTW and PS matching may be explained by the fact that the overlapped areas between PPV and SB were relatively small in the visualisation of propensity score distributions, suggesting a strong effect of the covariates on treatment selection (figures 2 and 3).

    Considering both the propensity score matching and IPTW with stabilised weight results, some of the patients who were treated by PPV in the current cohort would have lesser chances of reoperation if they had been treated by SB. Thus, PPV was not always superior to SB, contrary to the current trend, and SB was superior to PPV in some cases.


    This study is unique in that it controlled for recorded confounders pertaining to both patient characteristics and surgeon-related factors; however, there may be several unrecorded or unexplored confounders, such as the distance from the macula to the detached retina, existence of bullous RD and extent of vitreous liquefaction which might influence the choice of intervention and outcomes.

    Like other registry studies, credibility of data entry could not necessarily be guaranteed in this registry. Additionally, based on the surgeons’ preference, many surgical PPV interventions were accompanied by phacoemulsification, which might have affected the pure comparison between PPV and SB. Although RCTs are essential to conclude that one treatment is superior to the other, until well-designed clinical trials are available, we believe that registry-based research has an important role in providing some clues regarding an optimal treatment since circumstances for indications of the two treatments are changing due to progress in surgical technology.


    Although MIVS progress has contributed to an increase in the preference for PPV over SB for the treatment of RD, PPV may not be the optimal approach for repairing all types of RD. Our results revealed a lower rate of complications with SB for the treatment of simple phakic macula-on RD. Careful consideration is needed when selecting the right surgical technique to treat patients with simple phakic macula-on RD.

    Data availability statement

    No data are available.

    Ethics statements

    Patient consent for publication

    Ethics approval

    The main study protocol of the J-RD Registry was approved by the Ethics Committee of Kagoshima University (140093, 28-38), and the registry was compiled in accordance with the tenets of the Declaration of Helsinki, as well as with the Kyushu University Hospital Ethical Committee regulations.



    • Presented at Part of this data was presented at the XXXIth Meeting of the Club Jules Gonin, Jersey, Channel Islands, GB, and the Vail Vitrectomy, Vail, Colorado, USA, 2019.

    • Collaborators We would also like to thank the Institutions of the Japan Retinal Detachment Registry project committee and their collaborators: 1. Chiba University: Shuichi Yamamoto, Takayuki Baba, Eiju Sato, Masayasu Kitahashi, Tomoaki Tatsumi, Gen Miura and Tomohiro Niizawa. Kagoshima University: Taiji Sakamoto, Keita Yamakiri, Toshifumi Yamashita, Hiroki Otsuka, Seiji Sameshima, Narimasa Yoshinaga and Shozo Sonoda. Kyorin University: Akito Hirakata, Takashi Koto, Makoto Inoue, Kazunari Hirota, Yuji Itoh, Tadashi Orihara, Yoshinobu Emoto, Masahiko Sano, Hiroyuki Takahashi and Ryo Tokizawa. Yamagata University: Hidetoshi Yamashita, Koichi Nishitsuka, Yutaka Kaneko and Katsuhiro Nishi. Finally, we would like to thank the following collaborative institutions and collaborators: Asahikawa Medical University Hospital: Akitoshi Yoshida, Shinji Ono, Hiroyuki Hirokawa, Kenji Sogawa, Tsuneaki Omae and Akihiro Ishibazawa. Gunma University: Shoji Kishi, Hideo Akiyama, Hidetaka Matsumoto, Ryo Mukai and Masahiro Morimoto. Hirosaki University: Mitsuru Nakazawa, Yukihiko Suzuki, Takashi Kudo and Kobu Adachi. Hokkaido University: Susumu Ishida, Kousuke Noda, Satoru Kase, Syouhei Mori, Ryo Ando, Michiyuki Saito and Tomohiro Suzuki. Kansai Medical University Hospital: Kanji Takahashi, Yoshimi Nagai, Tadashi Nakauchi and Haruiko Yamada. Kindai University Sakai Hospital: Shuji Kusaka and Daishi Tsujioka. Kyoto University: Akitaka Tsujikawa and Kiyoshi Suzuma. Kyushu University: Tatsuro Ishibashi, Koh-Hei Sonoda, Yasuhiro Ikeda, Riichiro Kohno and Keijiro Ishikawa. Mie University: Mineo Kondo and Maki Kozawa. Nagasaki University: Takashi Kitaoka and Eiko Tsuiki. Nagoya City University: Yuichiro Ogura, Munenori Yoshida, Hiroshi Morita, Aki Kato, Yoshio Hirano and Kazuhiko Sugitani. Nagoya University: Hiroko Terasaki, Takeshi Iwase, Yasuki Ito, Shinji Ueno, Hiroki Kaneko, Norie Nonobe and Taro Kominami. National Center for Child Health and Development: Noriyuki Azuma and Tadashi Yokoi. Nihon University Hospital: Hiroyuki Shimada, Hiroyuki Nakashizuka, Takayuki Hattori, Ari Shinojima and Yorihisa Kutagawa. Okayama University: Fumio Shiraga, Yuki Morizane and Shuhei Kimura. Osaka Medical School: Tsunehiko Ikeda, Teruyo Kida, Takaki Sato and Masanori Fukumoto. Osaka Rosai Hospital: Kazuyuki Emi and Hiroshi Nakashima. Shiga Medical University: Masahito Ohji, Masashi Kakinoki and Osamu Sawada. Takeuchi Eye Clinic: Shinobu Takeuchi and Sumiyoshi Tanaka. Tokyo Womens Medical College: Tomohiro Iida, Hideki Koizumi, Ichiro Maruko, Taiji Hasegawa and Akiko Kogure. Yamanashi University: Hiroyuki Iijima, Tomohiro Oshiro, Yasushi Tateno, Wataru Kikushima, Atsushi Sugiyama and Seigo Yoneyama. Yokohama City University Medical Center: Kazuaki Kadonosono, Shimpei Sato and Shin Yamane.

    • Contributors SK and TI: study design; SK and TS: acquisition, data collection; SK, TI and TS: manuscript preparation; SK, TI and TS: manuscript revision; SK, TI and TS: overall responsibility.

    • Funding This work was supported by JSPS KAKENHI Grant Number JP18H02957.

    • Competing interests None declared.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    Linked Articles

    • At a glance
      Frank Larkin