Retina Specialists of Michigan

Uveal melanoma (UM) is an aggressive cancer of the eye that frequently results in metastatic death. UMs are most likely to metastasize when they are small, at a time when they are difficult to distinguish from benign nevi and often observed without treatment. Unfortunately, little is known about the early genetic evolution of UM or potential biomarkers to indicate small tumors undergoing malignant transformation. Here, we performed targeted next generation sequencing for the 7 canonical UM driver mutations in 1140 primary UMs, including 131 small early-stage tumors. We found that the evolutionary burst of genetic aberrations that determines the archetypal UM subtypes and metastatic propensity has already occurred by the time most small tumors are biopsied, although a significantly larger proportion of small tumors are still evolving compared to larger tumors. We found that the 15-gene expression profile (15-GEP) support vector machine discriminant score was the best indicator of tumors in transition from low-risk Class 1 to high-risk Class 2 signature. While BAP1, SF3B1 and EIF1AX mutations were associated with poor, intermediate and good prognosis, respectively, mutation analysis was inferior to the prospectively validated 15-GEP + PRAME expression classifier for predicting metastasis-free and overall survival. These results provide a more complete picture of genetic evolution in UM, and they move us closer to a molecular definition of malignant transformation in this cancer type.

Democratizing artificial intelligence (AI) enables model development by clinicians with a lack of coding expertise, powerful computing resources, and large, well-labeled data sets. To determine whether resource-constrained clinicians can use self-training via automated machine learning (ML) and public data sets to design high-performing diabetic retinopathy classification models. This diagnostic quality improvement study was conducted from January 1, 2021, to December 31, 2021. A self-training method without coding was used on 2 public data sets with retinal images from patients in France (Messidor-2 [n = 1748]) and the UK and US (EyePACS [n = 58 689]) and externally validated on 1 data set with retinal images from patients of a private Egyptian medical retina clinic (Egypt [n = 210]). An AI model was trained to classify referable diabetic retinopathy as an exemplar use case. Messidor-2 images were assigned adjudicated labels available on Kaggle; 4 images were deemed ungradable and excluded, leaving 1744 images. A total of 300 images randomly selected from the EyePACS data set were independently relabeled by 3 blinded retina specialists using the International Classification of Diabetic Retinopathy protocol for diabetic retinopathy grade and diabetic macular edema presence; 19 images were deemed ungradable, leaving 281 images. Data analysis was performed from February 1 to February 28, 2021. Using public data sets, a teacher model was trained with labeled images using supervised learning. Next, the resulting predictions, termed pseudolabels, were used on an unlabeled public data set. Finally, a student model was trained with the existing labeled images and the additional pseudolabeled images. The analyzed metrics for the models included the area under the receiver operating characteristic curve (AUROC), accuracy, sensitivity, specificity, and F1 score. The Fisher exact test was performed, and 2-tailed P values were calculated for failure case analysis. For the internal validation data sets, AUROC values for performance ranged from 0.886 to 0.939 for the teacher model and from 0.916 to 0.951 for the student model. For external validation of automated ML model performance, AUROC values and accuracy were 0.964 and 93.3% for the teacher model, 0.950 and 96.7% for the student model, and 0.890 and 94.3% for the manually coded bespoke model, respectively. These findings suggest that self-training using automated ML is an effective method to increase both model performance and generalizability while decreasing the need for costly expert labeling. This approach advances the democratization of AI by enabling clinicians without coding expertise or access to large, well-labeled private data sets to develop their own AI models.

To analyze the ophthalmic characteristics of congenital prepapillary vascular loop (PVL) and to propose a new morphologic classification dividing the loops into six types. Collaborative multinational multicentre retrospective study of PVL cases. There was a total of 49 cases (61 eyes), 37 unilateral (75.5%) and 12 bilateral (24.5%), 32 arterial type (65.3%) and 18 venous type (36.7%) (one patient had either kind in each eye). The mean number of loops per eye was 2.7 (range, 1-7). The loops were asymptomatic in 42 cases (85.7%). Other findings included: the presence of cilioretinal artery (14 cases), retinal vascular tortuosity (26 cases), amaurosis fugax (1 case), branch retinal artery occlusion (1 case) and vitreous haemorrhage (3 cases). Six morphologic loop types could be discerned based on elevation (flat vs. elevated), shape (figure of 8 or corkscrew with hyaline sheath), number (multiple or single), location (central or peripheral), lumen size (arterial vs. arteriolar) and presence of vascular tortuosity or vitreous traction. PVL are usually asymptomatic and can be divided into six morphologic types with different pathogenesis during early embryogenesis.

Prostate-specific antigen (PSA) screening may reduce death due to prostate cancer but leads to the overdiagnosis of many cases of indolent cancer. Targeted use of PSA screening may reduce overdiagnosis. Multimarker genomic testing shows promise for risk assessment and could be used to target PSA screening. To test whether counseling based on the family history (FH) and counseling based on a genetic risk score (GRS) plus FH would differentially affect subsequent PSA screening at 3 months (primary outcome), a randomized trial of FH versus GRS plus FH was conducted with 700 whites aged 40 to 49 years without prior PSA screening. Secondary outcomes included anxiety, recall, physician discussion at 3 months, and PSA screening at 3 years. Pictographs versus numeric presentations of genetic risk were also evaluated. At 3 months, no significant differences were observed in the rates of PSA screening between the FH arm (2.1%) and the GRS-FH arm (4.5% with GRS-FH vs. 2.1% with FH: χ2 = 3.13, P = .077); however, PSA screening rates at 3 months significantly increased with given risk in the GRS-FH arm (P = .013). Similar results were observed for discussions with physicians at 3 months and PSA screening at 3 years. Average anxiety levels decreased after the individual cancer risk was provided (P = .0007), with no differences between groups. Visual presentation by pictographs did not significantly alter comprehension or anxiety. This is likely the first randomized trial of multimarker genomic testing to report genomic targeting of cancer screening. This study found little evidence of concern about excess anxiety or overuse/underuse of PSA screening when multimarker genetic risks were provided to patients. Cancer 2016;122:3564-3575. © 2016 American Cancer Society.

9575 Background: Most uveal melanoma (UM) patients do not present with metastatic disease at diagnosis, yet 50% will develop metastases following primary tumor control. A validated 15-gene expression profile test enables risk-appropriate clinical management by using primary tumor tissue to predict metastatic risk, classifying patients as low risk Class 1A, intermediate risk Class 1B, or high risk Class 2. Methods: We performed an unplanned analysis of 70 patients in the prospective CLinical Application of DEcisionDx-UM Gene Expression Assay Results (CLEAR) registry (NCT02376920) because primary endpoints replicated in several prospective studies were met and accrual is terminating at the 4 participating centers. CLEAR aims at tracking clinical management and 5-year metastatic outcomes of Class 1 and 2 patients without evidence of distant metastasis at the time of primary tumor treatment. Surveillance modes and frequencies were independently determined by participating physicians as they deemed appropriate for Class 1 and 2 patients. Analysis cut-points paralleled a prior retrospective, multi-center decision impact study. Results: 37 (53%) patients had Class 1 results and 33 (47%) were Class 2. Overall, 12 Class 2 patients and 2 Class 1 patients had developed metastases (p = 0.002) with a median follow-up of 28.5 months (range, 2.6-94.0). At 3 years, metastasis-free survival rates were 100% for Class 1 and 62% for Class 2 (p = 0.002). All Class 2 patients and only 5/37 (14%) Class 1 patients were managed with high intensity surveillance (imaging and liver function tests every 3-6 months). This reflects a significant difference in management of Class 1 and Class 2 patients (p = 5.2x10-15). Only 4 Class 1 patients (11%) compared to 11 (33%) Class 2 patients were referred to medical oncology for surveillance and/or clinical trial enrollment (p = 0.04). Conclusions: Results support the prognostic accuracy of the 15-gene expression test for metastatic risk in UM patients. DecisionDx-UM risk classification impacts clinical decision-making as demonstrated by differences in frequency of imaging and bloodwork based on metastatic risk (higher intensity in Class 2 patients).

BackgroundPatients with type 2 diabetes (T2DM) often have poor glycemic control on first-line pharmacologic therapy and require treatment intensification. Intensification decisions can be difficult because of many available options and their many benefits and risks. The American Diabetes Association recommends patient-centered, evidence-based tools supporting shared decision-making between patients and clinicians. We developed a patient decision aid (PDA) targeting decisions about treatment intensification for T2DM. Our objective was to determine the effectiveness of this PDA for patients with T2DM on metformin who require treatment intensification.MethodsThis study was a pragmatic randomized controlled trial conducted in 27 US primary care and endocrinology clinics. Subjects were English-speaking adults with T2DM receiving metformin with persistent hyperglycemia who were recommended to consider medication intensification. Subjects were randomized to receive either the PDA or usual care (UC). Main outcome measures were change in knowledge, decisional self-efficacy, and decisional conflict.ResultsOf 225 subjects enrolled, 114 were randomized to the PDA and 111 to UC. Mean [SD] age was 52 [1] years, time since T2DM diagnosis was 6 [+/−6] years, 45.3 % were male, and most (55.5 %) were non-Caucasian. Compared to UC, PDA users had significantly larger knowledge gains (35.0 % [22.3] vs 9.9 % [22.2]; P < 0.0001) and larger improvements in self-efficacy (3.7 [16.7] vs−3.9 [19.2]; P < 0.0001) and decisional conflict (−22.2 [20.6] vs−7.5 [16.6]; P < 0.0001).ConclusionsThe PDA resulted in substantial and significant improvements in knowledge, decisional conflict and decisional self-efficacy. Decisional conflict scores after PDA use were within the range that correlates with effective decision-making. This PDA has the potential to facilitate shared-decision-making for patients with T2DM.Trial registration NCT02110979 Electronic supplementary materialThe online version of this article (doi:10.1186/s12913-016-1262-4) contains supplementary material, which is available to authorized users.

To estimate the duration of activity for intravitreal triamcinolone injected with a new technique using centrifuge concentration (Centrifuge concentrated IntraVitreal Triamcinolone, C-IVT). All injections were performed by a single surgeon (M.D.O.) using a 30-gauge needle. A vial of Triesence (triamcinolone; Alcon Laboratories, Fort Worth, TX) was drawn into a 1-mL syringe and the plunger cut off. The contents were spun in a centrifuge, and a second plunger was placed. Records of all patients receiving C-IVT with 0.05 mL or 0.1 mL from January 1, 2009, through December 31, 2009, were retrospectively reviewed. Eighty-four injections from 69 eyes of 57 patients were included. Sixty-nine injections from 54 eyes of 44 patients received 0.05 mL of C-IVT, whereas 15 injections from 15 eyes of 13 patients received 0.1 mL of C-IVT. Triamcinolone acetonide was still visualized in the vitreous on an average of 5.0 ± 2.4 months (median 5 months) after 0.05 mL of C-IVT and 8.3 ± 4.0 months (median 8 months) after 0.1 mL of C-IVT during follow-up visits. The longest duration recorded was 14 months for the 0.05-mL group and 18 months for the 0.l-mL group. The C-IVT results in a long duration of effect that seems to be greater than previously published techniques. It may be considered for patients requiring chronic steroid therapy, in which the benefits of long-term intravitreal steroids are believed to outweigh their risk.

Immunolabeling for the critical lymphocyte survival factor, Bcl-2, of intestinal biopsies from cats with histologic evidence of inflammatory bowel disease (IBD) or gastrointestinal (GI) lymphoma was evaluated to determine if expression differed significantly between these two disease processes. Immunolabeling for Bcl-2 was performed on small intestinal endoscopic or full thickness biopsy sections from 55 cats. Diagnosis of IBD, T-cell lymphoma or B-cell lymphoma was established previously. The percentage of infiltrating lymphocytes that were positively labeled for Bcl-2 was subjectively determined for each case. Eight cats were diagnosed with IBD and 47 cats with lymphoma. A significantly higher percentage of cells were positively immunolabeled for Bcl-2 in cats with GI lymphoma [median (range); 90 (5-95)%] compared with cats with IBD [60 (15-95)%] (P = 0.029). However, the overall degree of positive immunolabeling in both groups tended to be high. This over-expression of Bcl-2 may prove useful as a therapeutic target for IBD and GI lymphoma in cats.