Purpose
To investigate the potential of a combined assessment of clinical risk factors and biomarker profiling in the prediction of proliferative vitreoretinopathy (PVR) after retinal detachment surgery.
Design
Retrospective case-control study.
Methods
Multiplex bead-based immunoassays were used for the simultaneous measurement of 50 biomarkers in subretinal fluid samples obtained from patients who underwent scleral buckling surgery for primary rhegmatogenous retinal detachment (RRD). Of 306 samples that were collected and stored in our BioBank, we selected 21 samples from patients in whom a redetachment developed as a result of PVR within 3 months after reattachment surgery for primary RRD (PVR group). These were compared with age-, sex-, and storage time-matched RRD samples from 54 patients with an uncomplicated postoperative course after primary RRD repair (RRD group).
Results
Preoperative PVR was the only clinical variable that was an independent predictor of postoperative PVR development ( P = .035) and resulted in an area under the receiver operating characteristic curve of 0.67 (95% confidence interval, 0.51 to 0.83). The addition of the biomarkers chemokine (C-C motif) ligand 22, interleukin-3, and macrophage migration inhibitory factor improved the model significantly ( P < .001) and resulted in an area under the receiver operating characteristic curve of 0.93 (95% confidence interval, 0.82 to 1.04). A sensitivity of 94.1% and a specificity of 94.2% were reached, using a cutoff value of 5%.
Conclusions
In combination with preoperative PVR grade, the measurement of a single biomarker or a small multibiomarker panel shows great potential and may predict postoperative PVR development after primary RRD in a highly sensitive and specific manner.
Proliferative vitreoretinopathy (PVR) is a common cause of failure of reattachment surgery after primary rhegmatogenous retinal detachment (RRD). It is characterized by intraretinal fibrosis and the formation of cellular membranes on both sides of the retina. In fact, the PVR process is reminiscent of an aberrant wound-healing response in which several stages can be distinguished, including the influx of inflammatory cells, migration and proliferation of cells, and deposition and remodeling of extracellular matrix. To improve anatomic or visual success rates of reattachment strategies, research has focused on the use of intravitreal pharmacologic agents directed against one or more of these stages. So far, none of these is used in clinical practice on a routine basis because of lack of efficacy or concerns about retinal toxicity. For example, the combined intravitreal use of 5-fluorouracil and heparin did not show an improved outcome of vitreoretinal surgery for established PVR or unselected RRD cases. More promising results, however, were demonstrated for selected RRD patients at high risk for developing postoperative PVR. These studies have emphasized the need for the identification of high-risk subgroups to improve the risk-to-benefit ratio of pharmacologic adjunct strategies. If the development of postoperative PVR is predictable, only patients at high risk may be targeted with drugs that are potentially detrimental to retinal tissues. Moreover, these clinical trials have underscored the importance of research that focuses on the preclinical stages of PVR, rather than on patients with established PVR.
Although several studies have identified clinical risk factors for the development of PVR, attempts to evaluate the potential of biomarker profiling have been scarce until now. Because a wide range of cytokines has been suggested to play a role in the pathogenesis of PVR, it is likely that specific cytokines could serve as PVR biomarkers that can be used as prognostic factors in patients with primary RRD. Pathologic processes such as the breakdown of the blood-retinal barrier, the migration and proliferation of retinal pigment epithelial cells and glial cells, and the influx of inflammatory cells into the subretinal space may produce PVR-specific cytokine profiles. In previous studies on early biological alterations after primary RRD, we demonstrated a possible causal relationship between some cytokines and the development of postoperative PVR. In the present study, we extended the biomarker panel in the same patient population and addressed the potential of these biomarkers in a clinical setting. So far, prediction models have been constructed based on clinical parameters only. The aim of the present study was to identify biological markers that, in conjunction with clinical factors, are able to predict the future development of postoperative PVR with high sensitivity and high specificity.
Methods
Patients
Undiluted subretinal fluid samples were obtained from patients who underwent scleral buckling surgery for primary RRD. In our department, this surgical procedure was used only in patients with retinal detachments up to PVR grade C1 according to the 1983 classification. Subretinal fluid was not collected in relatively simple cases, that is, patients with small retinal detachments (< 1 quadrant involved) and patients with shallow detachments. Of 306 samples that were collected between 2001 and 2008, a redetachment developed in 45 patients as a result of PVR during the postoperative course that required repeat surgery, as was revealed by a thorough medical record study. Twenty-four of those were excluded for the following reasons: low sample volume or contamination with blood (n = 9), late PVR development (> 3 months after the primary surgical procedure; n = 6), preoperative vitreous hemorrhage (n = 4), preoperative trauma (n = 4), and preoperative cryotherapy (n = 1). None of the patients in whom postoperative PVR developed had preoperative uveitis. The remaining 21 samples comprised the PVR group, and these were compared with control subretinal fluid samples from the RRD group, that is, patients who had an uncomplicated follow-up after scleral buckling surgery for primary RRD. Samples from failures resulting from new breaks in the presence of PVR were not used as controls. Every single PVR sample was compared with 2 to 3 age-, sex-, and storage time-matched RRD samples, which resulted in 54 control samples. The same exclusion criteria applied to the RRD group.
Clinical Risk Factors for the Prediction of Proliferative Vitreoretinopathy
The following preoperative and intraoperative variables with the potential to induce or influence PVR development were collected retrospectively for all 75 patients: age, sex, size of retinal detachment, number of retinal defects, macular detachment, detachment duration, preoperative PVR grade, presence of diabetes, pseudophakia, intraoperative gas use, intraoperative cryotherapy, and intraoperative minor hemorrhage. The median follow-up time was 21 months in the PVR group (range, 3 to 80 months) and 6 months in the RRD group (range, 3 to 80 months). All patients were operated on by 3 experienced vitreoretinal surgeons (E.C.L.H., F.H., independent surgeon) who graded PVR according to the 1983 Classification of Retinal Detachment with PVR. To ensure consistency, they made fundus drawings of primary RRDs and recurrences, assessed signs of PVR, and, if PVR was present, assigned the appropriate PVR grade. For statistical purposes, data were collected as 0 in the absence of PVR, 1 for PVR grade A, 2 for PVR grade B, 3 for PVR grade C, and 4 for PVR grade D. Duration of retinal detachment was defined as the interval between the onset of symptoms and surgery and was estimated according to a precise history of patients’ symptoms. Duration of macular detachment was evaluated separately and was defined as the interval between the onset of a sudden drop in visual acuity and reattachment surgery.
Biomarkers for the Prediction of Proliferative Vitreoretinopathy
Biomarkers were measured in subretinal fluid samples obtained during scleral buckling surgery for primary RRD. Immediately after collection, undiluted subretinal fluid samples were stored at 4 degrees C. Within 15 minutes, samples were transferred to the BioBank Maastricht, where they were aliquoted in 50-μL portions and stored at −80 C until assayed, as described previously. Multiplex immunoassays (Luminex, Austin, Texas, USA) were performed at the Luminex Core Facility (Utrecht, The Netherlands) using an in-house–validated panel that incorporates an appropriate quality control program. In summary, the antibody-coated microspheres were incubated for 60 minutes with standards or subretinal fluid (50 μL). Plates were washed (Bio-Plex Pro II Wash Station; Bio-Rad, Hercules, California, USA), and a cocktail of biotinylated detection antibodies was added for an additional 60 minutes. After repeated washings, streptavidin–phycoerythrin was added and incubated for 10 minutes. Next, fluorescence intensity was measured and analysis of the data from all assays was performed (Bio-Plex System in combination with Bio-Plex Manager software version 4.1; Bio-Rad) using 5-parameter curve fitting. The concentrations of 50 biomarkers were measured ( Table 1 ). Concentrations above the upper detection limit were assigned the highest value from the respective standard curve, whereas concentrations below the lower detection limit were assigned the lowest value from the respective standard curve.
| Biological Group | Proteins |
|---|---|
| Interleukins | IL-1α, IL-1β, IL-2, IL-3, IL-4, IL-5, IL-6, IL-7, IL-9, IL-10, IL-11, IL-12p70, IL-13, IL-15, IL-17, IL-18, IL-21, IL-22, IL-23, IL-25, IL-33 |
| Growth factors | IGF-1, bFGF, NGF, HGF, VEGF |
| Chemokines | MIF, CCL2, CCL3, CCL5, CCL11, CCL17, CCL18, CCL19, CCL21, CCL22, CXCL8, CXCL9, CXCL10, CXCL12, CX3CL1 |
| Adhesion molecules | ICAM-1 |
| Adipokines | Adiponectin, leptin, chemerin, adipsin |
| Proteases/inhibitors | Cathepsin S, TIMP-1 |
| Others | TNF-α, IFN-γ |
Statistical Analysis
The outcome variable in the present study was postoperative PVR development (yes or no). Blockwise logistic regression analysis was used to identify clinical factors and biomarkers that were associated with this dichotomous outcome variable. In the first step (block 1), all clinical preoperative and intraoperative factors, regardless their P value, were considered as potential predictors for PVR development. In the second step (block 2), all biomarkers that reached P values < .05 in univariate logistic regression were considered as potential predictors for PVR development. Variables in both steps were entered with a forward procedure using a criterion of P < .10 for entering a variable in the model. The forward method for regression analysis was chosen because the backward method showed an unstable model as a result of collinearity between biomarkers. The ability of the constructed model to discriminate between those with a nonfavorable outcome (PVR group) and those with a favorable outcome (RRD group) was estimated by the area under the receiver operating characteristic curve (AUC). Cutoff values that yielded highest sensitivity and specificity were chosen. The individual risk of developing postoperative PVR after scleral buckling surgery for primary RRD was quantified in a prediction formula, in which the intercept was calibrated in accordance with the approximate incidence of postoperative PVR in our clinic and as reported in previous studies, that is, 10%. Analyses were performed using SPSS for Windows version 16.0 (SPSS, Inc, Chicago, Illinois, USA).
Results
Clinical Risk Factors for the Prediction of Proliferative Vitreoretinopathy
Clinical data and biomarker concentrations were available for all 75 included patients who underwent scleral buckling surgery for primary RRD. Twenty-one patients in whom a redetachment developed as a result of postoperative PVR were compared with 54 age-, sex-, and storage time-matched RRD control patients who had an uncomplicated postoperative course during the overall follow-up period. There were 15 men (71%) and 6 women (29%) in the PVR group, with a median age of 62 years (range, 43 to 76 years). The RRD group comprised 40 men (74%) and 14 women (26%), with a median age of 61 years (range, 43 to 79 years). The median storage time of the sample was 3.1 years in both groups. With respect to preoperative clinical parameters, there were no significant differences between the PVR group and the RRD group. Similar rates of intraoperative gas use (P = .807) and intraoperative cryotherapy (P = .585) were observed in both groups. The percentage of patients with minor intraoperative hemorrhage was higher significantly in patients in whom postoperative PVR developed (4/21 patients; 19%) as compared with those who had an uncomplicated postoperative course (1/54 patients; 2%; P = .007). Because this difference in baseline characteristics in this small number of patients may have jeopardized the robustness of the prediction model, we excluded these 5 cases from the blockwise logistic regression analysis. All preoperative and intraoperative clinical variables are summarized in Table 2 .
| Potential Clinical Risk Factor | RRD Group (n = 54) | PVR Group (n = 21) | Univariate Testing |
|---|---|---|---|
| Age (y) | |||
| Median (range) | 61 (43 to 79) | 62 (43 to 76) | NS |
| Sex (%) | |||
| Female | 26 | 29 | NS |
| Male | 74 | 71 | |
| Size of retinal detachment (quadrants) | |||
| Median (range) | 2 (1 to 3) | 2 (1 to 4) | NS |
| No. of retinal defects | |||
| Median (range) | 1 (0 to 7) | 1.5 (0 to 5) | NS |
| Macular detachment (%) | 64 | 86 | NS |
| Detachment duration (days) | |||
| Median (range) | 5 (1 to 75) | 6 (1 to 90) | NS |
| Preoperative PVR grade | |||
| Median (range) | 1 (0 to 3) | 2 (0 to 3) | NS |
| Diabetes mellitus (%) | 11 | 10 | NS |
| Pseudophakia (%) | 19 | 33 | NS |
| Intraoperative gas use (%) | 83 | 81 | NS |
| Intraoperative cryotherapy (%) | 65 | 71 | NS |
| Intraoperative minor hemorrhage (%) | 2 | 19 | P = .007 |
Biomarkers for the Prediction of Proliferative Vitreoretinopathy
Subretinal fluid levels of 50 different biomarkers, including interleukins (ILs), growth factors, chemokines, and adipokines were measured using multiplex immunoassays. Univariate logistic regression analysis showed that levels of IL-1α, IL-2, IL-3, IL-6, IL-11, macrophage migration inhibitory factor (MIF), chemokine (C-C motif) ligand 2 (CCL2), CCL3, CCL11, CCL17, CCL18, CCL19, CCL22, chemokine (C-X-C motif) ligand 10, cathepsin S, adiponectin, and intercellular adhesion molecule-1 were elevated significantly in the PVR group as compared with the RRD group, whereas levels of tissue inhibitor of metalloproteinase-1 were significantly lower (P < .05; Table 3 ).
| Biomarker | RRD Group (n = 54) | PVR Group (n = 21) | P Value (Univariate Testing) |
|---|---|---|---|
| IL-1α | 5.8 (2.9 to 11) | 7.0 (4.1 to 20) | .006 |
| IL-2 | 5.9 (3.7 to 12) | 6.6 (3.4 to 18) | .023 |
| IL-3 | 101 (24 to 348) | 144 (47 to 464) | .001 |
| IL-6 | 60 (8.3 to 1211) | 149 (36 to 2656) | .002 |
| IL-11 | 23 (7.3 to 59) | 25 (12 to 89) | .018 |
| MIF | 3618 (997 to 15 020) | 6691 (1691 to 12 900) | .008 |
| CCL2 | 849 (442 to 1139) | 930 (629 to 1134) | .044 |
| CCL3 | 359 (237 to 491) | 391 (268 to 662) | .035 |
| CCL11 | 8.58 (5.32 to 12.0) | 9.18 (6.90 to 15.0) | .001 |
| CCL17 | 2.04 (1.41 to 4.31) | 2.59 (1.76 to 5.84) | <.001 |
| CCL18 | 4305 (184 to 14 344) | 6897 (434 to 14 967) | .008 |
| CCL19 | 115 (21.4 to 705) | 309 (76.8 to 570) | <.001 |
| CCL22 | 18.2 (11.7 to 44.1) | 31.9 (10.9 to 92.5) | <.001 |
| CXCL10 | 206 (42.6 to 875) | 377 (128 to 1002) | .001 |
| TIMP-1 | 16 356 (8581 to > 26 200) | 14 401 (8724 to > 26 200) | .017 |
| Adiponectin | 17 699 (2300 to 38 737) | 26 877 (7810 to 46 132) | .002 |
| Cathepsin S | 4592 (1230 to 10 151) | 6119 (2326 to 9453) | .001 |
| ICAM-1 | 19 927 (4529 to 40 489) | 28 276 (6780 to 44 606) | .002 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
