To identify differences between Ranibizumab and Aflibercept in treatment-naïve patients with neovascular age-related macular degeneration (nvAMD) in a real-life clinical setting. We compared two groups of patients with a fairly similar prognosis either receiving Aflibercept or Ranibizumab within a pro re nata regimen for 1 year. Changes in visual acuity (letters) and central foveal thickness (CFT) and frequency of injections after completing the loading phase were evaluated using two separate multivariate mixed linear models. When correcting for baseline differences between the Aflibercept (11 eyes) and Ranibizumab (16 eyes) group, there was neither divergence in visual acuity (−0.97 letters (95 % CI. −6.06-4.12); p = 0.709), nor a significant difference in the reduction of CFT (−25.16 μm, 95 % CI; (−78.01-27.68); p = 0.351) between the two groups 1 year after treatment initiation. Also, the number of injection did not differ (0.04 (95 % CI; −0.16-0.09); p = 0.565). In contrast to health claims, treatment-naïve nvAMD, Ranibizumab and Aflibercept were equivalent in terms of functional and morphologic outcomes and number of injections when studied in real-life clinical practice.

Colour Doppler analysis of ophthalmic vessels has been proposed as a promising tool in the diagnosis of various eye diseases, but the available diagnostic evidence has not yet been assessed systematically. We performed a comprehensive systematic review of the literature on the diagnostic properties of Colour Doppler imaging (CDI) assessing ophthalmic vessels and provide an inventory of the available evidence. Eligible papers were searched electronically in (Pre) Medline, Embase and Scopus, and via cross-checking of reference lists. The minimum requirement to be included was the availability of original data and the possibility to construct a two-by-two table. Study selection, critical appraisal using the QUADAS II instrument and extraction of salient study characteristics was made in duplicate. Sensitivity and specificity was computed for each study. We included 11 studies (15 two-by-two tables) of moderate methodological quality enrolling 820 participants (range 30 to 118). In 44.4% participants were female (range 37–59% in specific subgroups). CDI was assessed for internal carotid stenosis, diabetic retinopathy, glaucoma, and branch or central retinal vein occlusion diagnosis. There was insufficient data to pool the results for specific illnesses. For the assessments of ophthalmic arteries, mean sensitivity was 0.69 (range 0.27–0.96) with a corresponding mean specificity of 0.83 (range 0.70–0.96). Mean sensitivity of the central retinal artery assessments was 0.58 (range 0.31–0.84) and the corresponding mean specificity was 0.82 (range 0.63–0.94). Robust assessments of the diagnostic value of colour Doppler analysis remain uncommon, limiting the possibilities to extrapolate its true potential for clinical practice. PROSPERO 2014:CRD42014014027.

Previous analyses of real-life data indicated that injection frequency and health care costs did not differ for anti-VEGF treatment with aflibercept and ranibizumab. The objective of this study was to investigate whether this finding persisted when analysing a longer time period after licensing. Retrospective analysis of health insurance claims data of two large Swiss basic health insurance plans including 28% of the Swiss population. Patients qualified for inclusion if aflibercept or ranibizumab treatment had been initiated between June 1, 2013 and November 1, 2014. Within this set, patients with at least 12 months of continuous insurance enrolment in the previous year, 12-month follow-up, and without change of anti-VEGF drug were considered. We examined the distribution of demographic data and patient characteristics between those receiving ranibizumab and those receiving aflibercept. Numbers of injections and associated health care expenditures observed during the 12-month follow-up period after incident treatment were the two outcomes considered. In multivariate regression analyses, controlling for possible confounding factors, we compared differences in these two outcomes between patients treated with aflibercept as compared to ranibizumab. A total of 3′058 patients were analysed, 790 (26%) receiving aflibercept and 2`268 receiving ranibizumab (74%). The use of aflibercept (average number of injections 6.2) as compared to ranibizumab (average number of injections 5.7) in the follow-up period of 1 to 12 months, was associated with a 12% increase in the injection frequency (95% confidence interval (CI) 6–17%; p < 0.001). Real-life data contradicts the assumption that aflibercept is used less frequently as compared to ranibizumab. This results in similar total health care expenditures for both anti-VEGF agents.

To compare anti-VEGF treatments for macular disease in terms of costs and clinical outcomes. We identified patients suffering from macular disease and treated either with aflibercept, ranibizumab or both at the largest public eye clinic in Switzerland between January 1st and December 31st 2016 who were insured in one of the two participating health insurance companies. Clinical data were extracted from the electronic health record system. The health insurers provided the health claim costs for the ophthalmologic care and the total health care costs of each patient in the observation period. Using multivariate regression models, we assessed the monthly ophthalmologic and the monthly total costs of patients with no history of switching (ranibizumab vs. aflibercept), patients with a history of switching from ranibizumab to aflibercept, patients switching during the observation period and a miscellaneous group. We examined baseline differences in age, proportion of males, visual acuity (letters), central retinal thickness (CRT) and treatment history before entering the study. We investigated treatment intensity and compared the changes in letters and CRT. The analysis involved 488 eyes (361 patients), 182 on ranibizumab treatment, and 63 on aflibercept treatment, 160 eyes with a history of switching from ranibizumab to aflibercept, and 45 switchers during follow-up and 38 eyes of the miscellaneous group. Compared to ranibizumab, monthly costs of ophthalmologic treatment were slightly higher for aflibercept treatment + 175.0 CHF (95%CI: 1.5 CHF to 348.3 CHF; p = 0.048) as were the total monthly costs + 581.0 CHF (95%CI: 159.5 CHF to 1002.4 CHF; p = 0.007). Compared to ranibizumab, the monthly treatment intensity with aflibercept was similar (+ 0.057 injections/month (95%CI -0.023 to 0.137; p = 0.162), corresponding to a projected annual number of 5.4 injections for ranibizumab vs. 6.1 injections for aflibercept. During follow-up, visus dropped by 0.7 letters with ranibizumab and increased by 0.6 letters with aflibercept (p = 0.243). CRT dropped by − 14.9 μm with ranibizumab and by − 19.5 μm with aflibercept (p = 0.708). The monthly costs of all other groups examined were higher. These real-life data show that aflibercept treatment is equally expensive, and clinical outcomes between the two drugs are similar.

Deep learning has huge potential to transform healthcare. However, significant expertise is required to train such models and this is a significant blocker for their translation into clinical practice. In this study, we therefore sought to evaluate the use of automated deep learning software to develop medical image diagnostic classifiers by healthcare professionals with limited coding – and no deep learning – expertise. We used five publicly available open-source datasets: (i) retinal fundus images (MESSIDOR); (ii) optical coherence tomography (OCT) images (Guangzhou Medical University/Shiley Eye Institute, Version 3); (iii) images of skin lesions (Human against Machine (HAM)10000) and (iv) both paediatric and adult chest X-ray (CXR) images (Guangzhou Medical University/Shiley Eye Institute, Version 3 and the National Institute of Health (NIH)14 dataset respectively) to separately feed into a neural architecture search framework that automatically developed a deep learning architecture to classify common diseases. Sensitivity (recall), specificity and positive predictive value (precision) were used to evaluate the diagnostic properties of the models. The discriminative performance was assessed using the area under the precision recall curve (AUPRC). In the case of the deep learning model developed on a subset of the HAM10000 dataset, we performed external validation using the Edinburgh Dermofit Library dataset. Diagnostic properties and discriminative performance from internal validations were high in the binary classification tasks (range: sensitivity of 73.3-97.0%, specificity of 67-100% and AUPRC of 0.87-1). In the multiple classification tasks, the diagnostic properties ranged from 38-100% for sensitivity and 67-100% for specificity. The discriminative performance in terms of AUPRC ranged from 0.57 to 1 in the five automated deep learning models. In an external validation using the Edinburgh Dermofit Library dataset, the automated deep learning model showed an AUPRC of 0.47, with a sensitivity of 49% and a positive predictive value of 52%. The quality of the open-access datasets used in this study (including the lack of information about patient flow and demographics) and the absence of measurement for precision, such as confidence intervals, constituted the major limitation of this study. All models, except for the automated deep learning model trained on the multi-label classification task of the NIH CXR14 dataset, showed comparable discriminative performance and diagnostic properties to state-of-the-art performing deep learning algorithms. The performance in the external validation study was low. The availability of automated deep learning may become a cornerstone for the democratization of sophisticated algorithmic modelling in healthcare as it allows the derivation of classification models without requiring a deep understanding of the mathematical, statistical and programming principles. Future studies should compare several application programming interfaces on thoroughly curated datasets.

The seminal work of Wennberg and Gittelsohn in 1973 emphasised the importance of health information for informed decision-making. This led to the creation of the Dartmouth Health Atlas in 1996, which has become an important resource for monitoring health services in the USA. The Dartmouth Health Atlas research revealed the existence of variation in health care without benefit to patients, and the dependence of health care use on local resource supply. Similar initiatives emerged around the world, from the UK to Asia. The availability of administrative data has become essential for evaluating health service delivery and for informing health economic analysis and policy decisions. Access to data depends on the organisation of the health system, with more centralised systems facilitating comprehensive data collection. We contrast the decentralised structure of the Swiss healthcare system with that of the US and the UK, and highlight the challenges of harmonising data for nationwide health monitoring. The example of optical coherence tomography (OCT) in Swiss ophthalmology illustrates the variability in care practices and billing patterns. This variability can be attributed to the lack of clear guidelines and the complexity of billing codes. Incentives to charge incorrect rates influence billing, adding a further variance component to the variance in care that cannot be subtracted from the total variance at the level of a health insurance fund and distorting the results. In certain environments the quality of data on care is so variable that a sound conclusions for health policy decisions represent a great challenge.