Literature Commentary

In this issue of Journal of Neuro-Ophthalmology, M. Tariq Bhatti, MD, and Mark L. Moster, MD will discuss the following 6 articles: 1. Powell G, Derry-Sumner H, Rajenderkumar D, Rushton SK, Sumner P. Persistent postural perceptual dizziness is on a spectrum in the general population. Neurology. 2020;94:e1929-e1938. 2. Mackie SL, Dejaco C, Appenzeller S, Camellino D, Duftner C, Gonzalez-Chiappe S, Mahr A, Mukhtyar C, Reynolds G, de Souza AWS, Brouwer E, Bukhari M, Buttgereit F, Byrne D, Cid MC, Cimmino M, Direskeneli H, Gilbert K, Kermani TA, Khan A, Lanyon P, Luqmani R, Mallen C, Mason JC, Matteson EL, Merkel PA, Mollan S, Neill L, Sullivan E, Sandovici M, Schmidt WA, Watts R, Whitlock M, Yacyshyn E, Ytterberg S, Dasgupta B. British Society for Rheumatology guideline on diagnosis and treatment of giant cell arteritis: executive summary. Rheumatology (Oxford). 2020;59:487-494. 3. Yang HK, Kim YJ, Sung JY, Kim DH, Kim KG, Hwang JM. Efficacy for differentiating nonglaucomatous vs glaucomatous optic neuropathy using deep learning systems. Am J Ophthalmol. [published online ahead of print April 2, 2020] doi:10.1016/j.ajo.2020.03.035. 4. Milea D, Najjar RP, Zhubo J, Ting D, Vasseneix C, Xu X, Fard MA, Fonseca P, Vanikieti K, Lagreze WA, Morgia CL, Cheung CY, Hamann S, Chiquet C, Sanda N, Yang H, Mejico LJ, Rougier MB, Kho R, Tran THC, Singhal S, Gohier P, Clermont-Vignal C, Cheng CY, Jonas JB, Yu-Wai-Man P, Fraser CL, Chen JJ, Ambika S, Miller NR, Liu Y, Newman NJ, Wong TY, Biousse V, the BONSAI Group. Artificial intelligence to detect papilledema from ocular fundus photographs. N Engl J Med. 2020;382:1687-1695. 5. Ghanem KG, Ram S, Rice PA. The modern epidemic of syphilis. N Engl J Med. 2020;382:845-854. 6. Woolen SA, Shankar PR, Gagnier JJ, MacEachern MP, Singer L, Davenport MS. Risk of nephrogenic systemic fibrosis in patients with stage 4 or 5 chronic kidney disease receiving a group II gadolinium-based contrast agent: a systematic review and meta-analysis. JAMA Intern Med. 2019;180:223-230.

Literature Commentary Section Editors: Mark L. Moster, MD M. Tariq Bhatti, MD Literature Commentary In this issue of Journal of Neuro-Ophthalmology, M. Tariq Bhatti, MD, and Mark L. Moster, MD will discuss the following 6 articles: 1. Powell G, Derry-Sumner H, Rajenderkumar D, Rushton SK, Sumner P. Persistent postural perceptual dizziness is on a spectrum in the general population. Neurology. 2020;94:e1929-e1938. 2. Mackie SL, Dejaco C, Appenzeller S, Camellino D, Duftner C, Gonzalez-Chiappe S, Mahr A, Mukhtyar C, Reynolds G, de Souza AWS, Brouwer E, Bukhari M, Buttgereit F, Byrne D, Cid MC, Cimmino M, Direskeneli H, Gilbert K, Kermani TA, Khan A, Lanyon P, Luqmani R, Mallen C, Mason JC, Matteson EL, Merkel PA, Mollan S, Neill L, Sullivan E, Sandovici M, Schmidt WA, Watts R, Whitlock M, Yacyshyn E, Ytterberg S, Dasgupta B. British Society for Rheumatology guideline on diagnosis and treatment of giant cell arteritis: executive summary. Rheumatology (Oxford). 2020;59:487-494. 3. Yang HK, Kim YJ, Sung JY, Kim DH, Kim KG, Hwang JM. Efficacy for differentiating nonglaucomatous vs glaucomatous optic neuropathy using deep learning systems. Am J Ophthalmol. [published online ahead of print April 2, 2020] doi:10.1016/j.ajo.2020.03.035. 4. Milea D, Najjar RP, Zhubo J, Ting D, Vasseneix C, Xu X, Fard MA, Fonseca P, Vanikieti K, Lagrèze WA, Morgia CL, Cheung CY, Hamann S, Chiquet C, Sanda N, Yang H, Mejico LJ, Rougier MB, Kho R, Tran THC, Singhal S, Gohier P, Clermont-Vignal C, Cheng CY, Jonas JB, Yu-Wai-Man P, Fraser CL, Chen JJ, Ambika S, Miller NR, Liu Y, Newman NJ, Wong TY, Biousse V, the BONSAI Group. Artificial intelligence to detect papilledema from ocular fundus photographs. N Engl J Med. 2020;382:1687-1695. 5. Ghanem KG, Ram S, Rice PA. The modern epidemic of syphilis. N Engl J Med. 2020;382:845-854. 6. Woolen SA, Shankar PR, Gagnier JJ, MacEachern MP, Singer L, Davenport MS. Risk of nephrogenic systemic fibrosis in patients with stage 4 or 5 chronic kidney disease receiving a group II gadolinium-based contrast agent: a systematic review and meta-analysis. JAMA Intern Med. 2019;180:223-230. Powell G, Derry-Sumner H, Rajenderkumar D, Rushton SK, Sumner P. Persistent postural perceptual dizziness is on a spectrum in the general population. Neurology. 2020;9:e1929-e1938 Objective: To examine the idea that symptoms of persistent postural perceptual dizziness (PPPD) are more common than previously assumed and lie on a spectrum in the general population, thus challenging current theories that PPPD is only a consequence of a vestibular insult. Methods: We collected 2 common clinical questionnaires of PPPD (the Visual Vertigo Analog Scale [VVAS] and Situational Characteristics Questionnaire [SCQ]) in the following 4 cohorts: community research volunteers (n = 1,941 for VVAS, n = 1,474 for SCQ); paid online participants (n = 190 for VVAS, n = 125 for SCQ); students (n = 204, VVAS only); and patients diagnosed with PPPD (n = 25). Results: We found that around 9%, 4%, and 11%, respectively, of the 3 nonclinical cohorts scored above the 25th percentile patient score on 1 PPPD measure (VVAS) and 49% and 54% scored above the 25th percentile patient score on the other measure (SCQ). Scores correlated negatively with age (counter to expectation). As expected, scores correlated with migraine in 2 populations, but this only explained a small part of the variance, suggesting that migraine is not the major factor underlying the spectrum of PPPD symptoms in the general population. 434 Conclusion: We found high levels of PPPD symptoms in nonclinical populations, suggesting that PPPD is a spectrum that preexists in the population, rather than only being a consequence of vestibular insult. Atypical visuo-vestibular processing predisposes some individuals to visually induced dizziness, which is then exacerbated should a vestibular insult (or more generalized insult) occur. COMMENTS Since coming to the Mayo Clinic, I have seen quite a few patients with PPPD probably because Dr. Jeffrey P. Staab, who is at the Mayo Clinic and one of the world's top thought leaders in this disorder, has attracted many patients to seek consultation from him and his team. In fact, I must admit before 2 years ago, I was not familiar with PPPD. I am not ashamed to say I looked up the diagnostic criteria for PPPD, a chronic vestibular disorder, and they are as follows: 1. Dizziness, unsteadiness, or nonspinning vertigo lasting for hours at a time, present on most days for 3 months or more. 2. Persistent spontaneous symptoms exacerbated by upright posture, active or passive motion, and moving visual stimuli or visual patterns. Moster and Bhatti: J Neuro-Ophthalmol 2020; 40: 434-441 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Literature Commentary TABLE 1. Results General population Paid participants Undergraduate psychology students At or Above 25th Percentile of the VVAS Patient Score, n% Above the Minimum VVAS Patient Score, n% 9 4 11 13 9 23 At or Above the 25th Percentile of the SCQ Patient Score 49% 54% Did not take questionnaire Above the Minimum SCQ Patient Score 83% 86% Did not take questionnaire SCQ, situational characteristics questionnaire; VVAS, visual vertigo analog scale. 3. Precipitated by causing conditions that cause vertigo, unsteadiness, or dizziness, such as peripheral or central vestibular disorders, vestibular migraine, pain attacks, anxiety, head trauma, and autonomic disorders. 4. Symptoms causing significant distress or functional impairment. 5. Symptoms not better accounted for by another disease but can co-exist with PPD. This 4-questionnaire study was conducted to determine how many people in the general population have symptoms that could be considered part of PPPD. The following four cohorts were created: • General population (n = 1,941 for the VVAS, n = 1,474 for the SCQ). Average age was 55 years; 74% were women. • Paid participants (n = 211). Average age was 27 years; 30% were women. • Undergraduate psychology students (n = 204). Average age was 19 years; 85% were women. • Patients with PPPD (n = 25). Average age was 44 years; 60% were women. The 2 main questionnaires were the VVAS and SCQ, but subjects were also asked to complete a migraine screening questionnaire and hospital anxiety and depression scale questionnaire. Any subject with a self-reported history of a vestibular disorder was excluded, but vestibular migraine was not an exclusion criterion. The results are shown in Table 1. • Individuals with greater number of PPPD symptoms were more likely to have experienced migraine headaches • There was a positive correlation of increased PPPD symptoms with anxiety and depression The following two hypotheses were offered by the authors to explain their findings: participants may have had a previous vestibular problem and varying visual dependence of people. They suggested ".that atypical visuo-vestibular processing predisposes some individuals to visually induced dizziness, which is then exacerbated should a vestibular insult (or more generalized insult) occur." Moster and Bhatti: J Neuro-Ophthalmol 2020; 40: 434-441 The take-home points of this study were that the symptoms of PPPD are common in the general population (i.e., subclinical), PPPD is not a categorical diagnosis, and has a variable severity profile (similar to what we discussed in the last issue of Literature Commentary regarding visual snow syndrome (1)). What this study tells me, and something that can be said for many other diseases I think, is that there is a background of symptoms many of us carry that do not become clinically apparent and are considered "normal". -M. Tariq Bhatti, MD 1. Puledda F, Schankin C, Goadsby PJ. Visual snow syndrome: a clinical and phenotypical description of 1,100 cases. Neurology. 2020;94:e564-e574. Tariq, as I've pointed out in the past, there are books written about what you haven't heard of. I'm glad you moved to the Mayo Clinic where you are being educated. PPPD is not a new entity but a new name for what used to be called "phobic postural vertigo," "chronic subjective dizziness," and a few other names. It's interesting that PPPD is described in this article as "functional," something we think of as nonorganic, but the article defining the criteria for PPPD published by Dr. Staab (1) says that it is not psychiatric. To quote from that article, ".functional is not a synonym for psychogenic or psychosomatic as it was throughout most of the 20th century, and therefore, does not reflect a presumption of psychopathological abnormalities." Despite this claim, it has been associated with anxiety, depression, and other psychological conditions. This article suggests that many people in the normal population are predisposed to PPPD and having had a previous vestibular insult helps unmask it. -Mark L. Moster, MD 1. Staab JP, Eckhardt-Henn A, Horii A, Jacob R, Strupp M, Brandt T, Bronstein A. Diagnostic criteria for persistent posturalperceptual dizziness (PPPD): consensus document of the committee for the Classification of Vestibular Disorders of the Bárány Society. J Vestib Res. 2017;27:191-208. 435 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Literature Commentary No wonder I hadn't heard of PPPD, "What's in a name? That which we call PPPD by any other name would spin as sweet." -M. Tariq Bhatti, MD 2. Mackie SL, Dejaco C, Appenzeller S, Camellino D, Duftner C, Gonzalez-Chiappe S, Mahr A, Mukhtyar C, Reynolds G, de Souza AWS, Brouwer E, Bukhari M, Buttgereit F, Byrne D, Cid MC, Cimmino M, Direskeneli H, Gilbert K, Kermani TA, Khan A, Lanyon P, Luqmani R, Mallen C, Mason JC, Matteson EL, Merkel PA, Mollan S, Neill L, Sullivan E, Sandovici M, Schmidt WA, Watts R, Whitlock M, Yacyshyn E, Ytterberg S, Dasgupta B. British Society for Rheumatology guideline on diagnosis and treatment of giant cell arteritis: executive summary. Rheumatology (Oxford). 2020;59:487-494 3. 4. 5. 6. 7. No abstract available 8. COMMENTS Because "recent years have seen new evidence emerge regarding the diagnosis and treatment of giant cell arteritis (GCA)", the British Society of Rheumatology (BSR) decided to update their 2010 guidelines. The 2020 BSR GCA guidelines were developed by a group of health care providers that included rheumatology, general medicine, and ophthalmology. There were 3 broad categories of GCA diagnosis and treatment as follows: general principles, diagnostic tests, and treatment. Recommendations were based on the Grading of Recommendations, Assesment, Development and Evaluation (GRADE) methodology (strong and conditional), determined by the quality of evidence. Each recommendation received a consensus score from 0 to 10. The quality of evidence was assessed as follows: • High: +++++ (A) • Moderate: ++++ (B) • Low: +++ (C) • Very low: ++ (D) The executive summary provides the information in a very concise matter, but for those interested the full guidelines were also published (1). Many of the recommendations are things we are supposed to and already do, but I wanted to highlight a few of the following guidelines: 1. Baseline investigations: C-reactive protein (CRP) for all patients with either erythrocyte sedimentation rate (ESR) or plasma viscosity and complete blood count; consider serum and urine protein electrophoresis if ESR elevated out of proportion to CRP; plasma glucose, renal and liver function tests, calcium and alkaline phosphatase; urine 436 9. 10. 11. 12. dipstick, chest radiograph, tuberculosis testing; thyroidstimulating hormone, vitamin D, bone density test, and dual-energy X-ray absorptiometry. Strong recommendation of a confirmatory test is either with a temporal artery biopsy (TAB) (at least 1 cm length) or ultrasound of the temporal and axillary arteries (or both). Quality of Evidence (QOE): +++, consensus score: 9.33. In a patient with a high probability of GCA, a positive ultrasound is all that is needed to confirm the diagnosis. In a patient with a low probability of GCA, a negative ultrasound is all that is needed to exclude the diagnosis. It was noted that the TAB had relatively greater rule-in value, and ultrasound had a relatively greater rule-out value. A couple of caveats to the ultrasound were as follows: the halo sign decreases in size during the first week of steroid treatment and the results are operatordependent requiring "adequate" training. Conditional recommendation of a standard initial dose of glucocorticoids is 40-60 mg/day of oral prednisone. QoE: +, consensus score: 9.44. Conditional recommendation of glucocorticoid treatment for acute or intermittent visual loss can be 500 mg or 1 g intravenous methylprednisolone daily for 3 consecutive days. QoE: +, consensus score: 9.00. Conditional recommendation of glucocorticoid taper to zero over 12-18 months. QoE: +, consensus score: 8.81. Conditional recommendation of combining methotrexate with glucocorticoid in patients at a high risk of glucocorticoid toxicity or who relapse. QoE: ++, consensus score: 8.92. Strong recommendation of adding tocilizumab during glucocorticoid taper in patients at a high risk of glucocorticoid toxicity or who relapse. QoE: +++, consensus score: 9.61. Routine use of antiplatelet or anticoagulation therapy not recommended. I wanted to review these guidelines because to my surprise the American College of Rheumatology, which published classification criteria for GCA in 1990 (2), has not published any such clinical guidelines to date. -M. Tariq Bhatti, MD 1. MackieSL, DejacoC, AppenzellerS, CamellinoD, DuftnerC, Gonzalez-ChiappeS, MahrA, MukhtyarC, ReynoldsG, de SouzaAWS, BrouwerE, BukhariM, ButtgereitF, ByrneD, CidMC, CimminoM, DireskeneliH, GilbertK, KermaniTA, KhanA, LanyonP, LuqmaniR, MallenC, MasonJC, MattesonEL, MerkelPA, MollanS, NeillL, SullivanE, SandoviciM, SchmidtWA, WattsR, WhitlockM, YacyshynE, YtterbergS, DasguptaB. British Society for Rheumatology guideline on diagnosis and treatment of giant cell arteritis. Rheumatology (Oxford). 2020;59:e1-e23. 2. HunderGG, BlochDA, MichelBA, StevensMB, ArendWP, CalabreseLH, EdworthySM, FauciAS, LeavittRY, LieJT, Moster and Bhatti: J Neuro-Ophthalmol 2020; 40: 434-441 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Literature Commentary LightfootRW, MasiAT, McShaneDJ, MillsJA, WallaceSL, ZvaiflerNJ. The American College of Rheumatology 1990 criteria for the classification of giant cell arteritis. Arthritis Rheum. 1990;33:1122-1128. Tariq, to quote David Brinkley "Everyone is entitled to my opinion!" My opinion is that the most expert clinicians for GCA are rheumatologists and neuro-ophthalmologists. This panel of 36 specialists seems to have one comprehensive ophthalmologist (Eoin O' Sullivan) and although I don't know all neuro-ophthalmologists in the world (Tariq, who?), I think there is only one neuro-ophthalmologist on this panel (Susan Mollan). Some of the BSR recommendations are different than what many of us practice. For instance, they rely strongly on arterial ultrasound and somewhat strongly on MRI changes in vessels, which in our practice has been unhelpful. Although I have been against using ultrasound for diagnosing GCA, I have to admit their approach makes sense in centers with sufficient experience and expertise with ultrasound for this purpose. To reiterate, the way they use it is as follows: 1. If symptoms, signs, and laboratory tests suggest a low probability of GCA and ultrasound is negative, consider alternative diagnoses. If ultrasound is positive, then perform a TAB. 2. If symptoms, signs, and laboratory tests suggest a high probability of GCA and ultrasound is positive, treat for GCA. If ultrasound is negative, perform a TAB. 3. If symptoms, signs, and laboratory tests suggest a medium probability of GCA or ultrasound is equivocal, perform a TAB. I have to say that even to someone like me who always insists on a TAB, this approach in the appropriate institution is quite reasonable. -Mark L. Moster, MD Yang HK, Kim YJ, Sung JY, Kim DH, Kim KG, Hwang JM. Efficacy for differentiating nonglaucomatous vs glaucomatous optic neuropathy using deep learning systems. Am J Ophthalmol. [published online ahead of print April 2, 2020] doi:10.1016/ j.ajo.2020.03.035 Purpose: To assess the performance of deep learning approaches for differentiating nonglaucomatous optic neuropathy (NGON) vs glaucomatous optic neuropathy (GON) on color fundus photographs by the use of image recognition. Design: Development of an Artificial Intelligence Classification algorithm METHODS: Setting: Institutional. Subjects: An analysis including 3,815 fundus images from the Picture Archiving and Communications system of Seoul National University Bundang Hospital consisting of 2,883 normal optic disc images, 446 NGON with optic disc pallor, and 486 GON. Moster and Bhatti: J Neuro-Ophthalmol 2020; 40: 434-441 Observations: The presence of NGON and GON was interpreted by 2 expert neuro-ophthalmologists and had corroborate evidence on visual field testing and optical coherence tomography. Images were preprocessed in size and color enhancement before input. We applied the convolutional neural network of ResNet-50 architecture. The area under the precision-recall curve (average precision [AP]) was evaluated for the efficacy of deep learning algorithms to assess the performance of classifying nonglaucomatous optic disc pallor and GON. Results: The diagnostic accuracy of the ResNet-50 model to detect GON among NGON images showed a sensitivity of 93.4% and specificity of 81.8%. The area under the precision-recall curve for differentiating NGON vs GON showed an AP value of 0.874. False positive cases were found with extensive areas of peripapillary atrophy and tilted optic discs. Conclusion: Artificial intelligence-based deep learning algorithms for detecting optic disc diseases showed excellent performance in differentiating NGON and GON on color fundus photographs, necessitating further research for clinical application. COMMENTS No doubt that differentiating nonglaucomatous cupping from glaucomatous cupping can be challenging and I want to acknowledge the excellent review article by Dr. Fraser et al (1) on this subject matter published in JNO in 2013. This study used the development of an artificial intelligence (AI) classification algorithm to detect nonglaucomatous cupping from glaucomatous cupping. The machine analyzed 3,815 fundus images of which 2,883 were normal optic discs, 446 nonglaucomatous cupping with disc pallor, and 486 glaucomatous cupping. The deep learning of the machine was performed based on 900 images (300 normal optic discs, 300 nonglaucomatous cupping, and 300 glaucomatous cupping) for the training set, 240 images (80 normal optic discs, 80 nonglaucomatous cupping, and 80 glaucomatous cupping) for the validation set, and 2,675 images (2,503 normal optic discs, 66 nonglaucomatous cupping, and 106 glaucomatous cupping) for the test set. Because of insufficient number of data in the training set, it was augmented by 20-fold by what the authors state as "random combination of flip, rotation, translation, and scale adjustment." The overall accuracy of the machine was 99.1% (normal optic discs 99.7%, nonglaucomatous cupping 86.4%, and glaucomatous cupping 92.5%). The sensitivity and specificity to detect nonglaucomatous cupping from normal optic disc and glaucomatous cupping was 86.4% and 99.6%, respectively. The sensitivity and specificity to detect glaucomatous cupping from nonglaucomatous cupping was 93.4% and 81.8%, respectively, with the area under the precision-recall curve being 0.874. The 14 false-positive results for the glaucomatous cupping group were due to the presence of peripapillary atrophy and tilted optic discs. The 437 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Literature Commentary 7 false-positive results in the nonglaucomatous cupping were due to peripapillary atrophy and dark background illumination. The authors listed several limitations to their study, including, patients with variable glaucomatous and nonglaucomatous cupping, variable brightness and image quality of the photographs, and unequal proportion of images in the 3 groups (the reason for using the precisionrecall curve rather than the receiver operating curve), supplementation of the training set data, image duplication bias, and lack of ethnic diversity among the photographs. I am not sure if this system is ready for prime time just yet because it needs more validation especially as it applies to different ethnicities. Also, I would like to have seen the authors challenge the AI against several neuroophthalmologists. Do you think you would win Mark? -M. Tariq Bhatti, MD 1. FraserCL, WhiteAJ, PlantGT, MartinKR. Optic nerve cupping and the neuro-ophthalmologist. J Neuroophthalmol. 2013;33:377-389. The use of AI for fundus imaging will likely improve our ability to make more accurate diagnoses in the future. I think it is very good that in the deep learning phase the nonglaucomatous cases were not just based on the neuroophthalmologists' view of the image but on corroborative findings on visual fields and ocular coherence tomography and these patients had workups and diagnoses obtained. Even so, the program did not do accurately enough with the nonglaucomatous cases. To answer your question, I don't think I would win against AI, but I'm also glad not to win when I think of what Steven Hawking said: "Artificial intelligence is likely to be either the best or the worst thing to happen to humanity!" -Mark L. Moster, MD Milea D, Najjar RP, Zhubo J, Ting D, Vasseneix C, Xu X, Fard MA, Fonseca P, Vanikieti K, Lagrèze WA, Morgia CL, Cheung CY, Hamann S, Chiquet C, Sanda N, Yang H, Mejico LJ, Rougier MB, Kho R, Tran THC, Singhal S, Gohier P, Clermont-Vignal C, Cheng CY, Jonas JB, Yu-Wai-Man P, Fraser CL, Chen JJ, Ambika S, Miller NR, Liu Y, Newman NJ, Wong TY, Biousse V; the BONSAI Group. Artificial intelligence to detect papilledema from ocular fundus photographs. N Engl J Med. 2020;382(18):1687-1695 Background: Nonophthalmologist physicians do not confidently perform direct ophthalmoscopy. The use of artificial intelligence to detect papilledema and other optic disc abnormalities from fundus photographs has not been well studied. 438 Methods: We trained, validated, and externally tested a deep-learning system to classify optic discs as being normal or having papilledema or other abnormalities from 15,846 retrospectively collected ocular fundus photographs that had been obtained with pharmacologic pupillary dilation and various digital cameras in persons from multiple ethnic populations. Of these photographs, 14,341 from 19 sites in 11 countries were used for training and validation, and 1,505 photographs from 5 other sites were used for external testing. Performance at classifying the optic disc appearance was evaluated by calculating the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity, as compared with a reference standard of clinical diagnoses by neuro-ophthalmologists. Results: The training and validation data sets from 6,779 patients included 14,341 photographs: 9,156 of normal discs, 2,148 of discs with papilledema, and 3,037 of discs with other abnormalities. The percentage classified as being normal ranged across sites from 9.8% to 100%; the percentage classified as having papilledema ranged across sites from zero to 59.5%. In the validation set, the system discriminated disks with papilledema from normal disks and disks with nonpapilledema abnormalities with an AUC of 0.99 (95% confidence interval [CI], 0.98-0.99) and normal from abnormal disks with an AUC of 0.99 (95% CI, 0.99- 0.99). In the external-testing data set of 1,505 photographs, the system had an AUC for the detection of papilledema of 0.96 (95% CI, 0.95-0.97), a sensitivity of 96.4% (95% CI, 93.9-98.3), and a specificity of 84.7% (95% CI, 82.3-87.1). Conclusions: A deep-learning system using fundus photographs with pharmacologically dilated pupils differentiated among optic discs with papilledema, normal discs, and discs with nonpapilledema abnormalities. Funded by the Singapore National Medical Research Council and the SingHealth Duke-NUS Ophthalmology and Visual Sciences Academic Clinical Program. COMMENTS Because you brought up artificial intelligence in the last article, let's continue with this article. This is a terrific article by Don Milea and many other colleagues of ours studying whether a deep-learning system could aid in the diagnosis of papilledema from fundus photographs. The study used a very large number of photographs (14,341) to train and validate the deep-learning system, including 2,148 with papilledema and 3,037 with other abnormalities including nonarteritic anterior ischemic optic neuropathy, optic neuritis, optic disc drusen, optic atrophy, and congenital optic nerve abnormalities. It then tested the system externally on 1,505 photographs from 5 centers. The main finding was that the system could discriminate among normal optic discs, discs with papilledema, and discs with other abnormalities. The sensitivity for detecting papilledema was 96.4% and the specificity was 84.7%. Negative predictive values were 99.6%, but positive predictive value was only 39.8% and varied with the prevalence of papilledema in the 5 different centers. Moster and Bhatti: J Neuro-Ophthalmol 2020; 40: 434-441 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Literature Commentary Fifteen of 360 (4.2%) photographs of optic discs with papilledema were missed. They were classified as other abnormalities, but none were classified as normal. The authors note numerous limitations, including the threshold for diagnosing papilledema was set low to avoid false negatives. Although the goal of such research is to help nonexperts diagnose optic nerve disease, I think it will be quite helpful to neuro-ophthalmologists as well. When AI findings agree with my clinical suspicion of papilledema or pseudopapilledema, I will be reassured and sleep better at night. When the findings conflict with my clinical opinion, I will be glad to reconsider my opinion. I imagine many in our field will predict that this will decrease referrals to neuro-ophthalmologists as many predicted when MRI came out. However, we have seen MRI abnormalities increase referrals. In addition, if referrals do go down, the ones we do see may be more appropriate, which could be a benefit in view of the shortage of neuroophthalmologists that exists. -Mark L. Moster, MD I agree with you Mark, this is a terrific study! What I would like to see is how AI compares to a neuroophthalmologist in real time. -M. Tariq Bhatti, MD Ghanem KG, Ram S, Rice PA. The modern epidemic of syphilis. N Engl J Med. 2020;382(9):845-854. No Abstract available COMMENTS Because every journal is filled with COVID-19 articles, I decided to remind us about a different infection very near and dear to neuro-ophthalmology by reporting on this review article about syphilis. Of major concern is an 81% increase in cases between 2014 and 2018 in the United States. Eighty-six percent of patients are men. More than half of the men have sex with other men and 42% also have HIV. In addition, the risk in women increases 6-fold if using intravenous (IV) drugs or having sex with an IV drug user. In neuro-ophthalmology practice, we typically order both nontreponemal tests (NTTs) (e.g., rapid plasma reagin and venereal disease research laboratory tests) and treponemal tests (TTs) (e.g., fluorescent treponemal antibody and treponema pallidum antibody tests) for the diagnosis of ocular or neurosyphilis. This article talks about 2 approaches to testing, starting either with "either" a tradiMoster and Bhatti: J Neuro-Ophthalmol 2020; 40: 434-441 tional approach with the NTTs or a "reverse sequence" starting with the TTs. With the traditional approach, if the NTT is reactive and then a TT is negative, it is a biologic false positive. If NTT is reactive and TT is reactive it is consistent with active or treated syphilis. If NTT is not reactive then it is most likely not syphilis. However, very early syphilis and long-standing treated or untreated syphilis can also be nonreactive. With the reverse sequence approach, if the TT is reactive and then an NTT is reactive, the diagnosis is syphilis. If the TT is reactive and then NTT is negative a confirmatory TT is performed. If reactive, then treated syphilis is most likely, but long-standing untreated or early primary syphilis is possible. If the confirmatory TT is negative, it is likely a biologic false positive. If the TT is negative, it is unlikely to be syphilis but rarely can be early or long-standing treated syphilis. Although the reverse sequence approach makes sense in patients where the clinical picture is beyond the primary syphilis stage, my experience is that our laboratories often will not do the TTs unless I first find a positive NTT. For treatment, they quote current Center for Disease Control guidelines for neurosyphilis and ocular syphilis as follows: 1. Aqueous crystalline penicillin G, 18-24 million units per day, administered in IV doses of 3-4 million units every 4 hours or as a continuous infusion, for 10-14 days or 2. Penicillin G procaine, 2.4 million units in a single intramuscular dose daily, plus probenecid, 500 mg administered orally 4 times a day, both for 10-14 days (alternative). For penicillin allergy they recommend desensitization but point out that ceftriaxone penetrates the central nervous system well and is an option for treating neurosyphilis. For ocular syphilis, they claim a cerebrospinal fluid examination is not necessary if serologic tests are positive because up to 30% are negative. This review article does not contribute much that is new, but points out the recent epidemic and provides a nice review. -Mark L. Moster, MD Thank you for reviewing this article Mark, it's always nice to revisit a disease that we think we know all about but may have forgotten some of the subtleties. In addition to your points, I just want to highlight a few points regarding ocular syphilis mentioned in the article. Ocular syphilis is a separate manifestation of neurosyphilis; in other words, the central nervous system involvement is not a necessary outcome for the development of ocular syphilis but both can occur concomitantly. In addition, ocular syphilis can occur during any of the 3 stages of infection. -M. Tariq Bhatti, MD 439 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Literature Commentary Woolen SA, Shankar PR, Gagnier JJ, MacEachern MP, Singer L, Davenport MS. Risk of nephrogenic systemic fibrosis in patients with stage 4 or 5 chronic kidney disease receiving a group II gadolinium-based contrast agent: a systematic review and meta-analysis. JAMA Intern Med. 2019;180:223-230. Importance: Risk of nephrogenic systemic fibrosis (NSF) to individual patients with Stage 4 or 5 chronic kidney disease (CKD; defined as estimated glomerular filtration rate [GFR] of ,30 mL/min/1.73 m2) who receive a group II gadolinium-based contrast agent (GBCA) is not well understood or summarized in the literature. Objective: To assess the pooled risk of NSF in patients with Stage 4 or 5 CKD receiving a group II GBCA. Data Sources: A health sciences informationist searched the Ovid (MEDLINE and MEDLINE ePub Ahead of Print, InProcess & Other Non-Indexed Citation, and Daily and Versions), Embase, Cochrane Central Register of Controlled Trials, Web of Science, and OpenGrey databases from inception to January 29, 2019, yielding 2,700 citations. Study Selection: Citations were screened for inclusion in a multistep process. Agreement for final cohort inclusion was determined by 2 blinded screeners using Cohen k. Inclusion criteria consisted of Stage 4 or 5 CKD with or without dialysis, administration of an unconfounded American College of Radiology (ACR) classification group II GBCA (gadobenate dimeglumine, gadobutrol, gadoterate meglumine, or gadoteridol), and incident NSF as an outcome. Conference abstracts, retracted manuscripts, narrative reviews, editorials, case reports, and manuscripts not reporting total group II GBCA administrations were excluded. Data Extraction and Synthesis: Data extraction was performed for all studies by a single investigator, including publication details, study design and time frame, patient characteristics, group II GBCA(s) administered, total exposures for patients with Stage 4 or Stage 5 CKD, total cases of unconfounded NSF, reason for GBCA administration, follow-up duration, loss to follow-up, basis for NSF screening, and diagnosis. Main Outcomes and Measures: Pooled incidence of NSF and the associated upper bound of a 2-sided 95% CI (risk estimate) for the pooled data and each of the 4 group II GBCAs. Results: Sixteen unique studies with 4,931 patients were included (k = 0.68) in this systematic review and metaanalysis. The pooled incidence of NSF was 0 of 4,931 (0%; upper bound of 95% CI, 0.07%). The upper bound varied owing to different sample sizes for gadobenate dimeglumine (0 of 3,167; upper bound of 95% CI, 0.12%), gadoterate meglumine (0 of 1,204; upper bound of 95% CI, 0.31%), gadobutrol (0 of 330; upper bound of 95% CI, 1.11%), and gadoteridol (0 of 230; upper bound of 95% CI, 1.59%). Conclusions and Relevance: This study's findings suggest that the risk of NSF from group II GBCA administration in Stage 4 or 5 CKD is likely less than 0.07%. The potential diagnostic harms of withholding group II GBCA for indicated 440 examinations may outweigh the risk of NSF in this population. COMMENTS Many patients with neuro-ophthalmologic presentations have coexisting kidney disease. Not being able to give a gadolinium agent when performing MRI on these patients has been an impediment to accurate diagnosis in many cases. This occurred after reports of more than 500 cases of NSF resulted in a black box warning by the Food and Drug Administration advising against any gadolinium contrast agent in patients with a GFR of less than 30 mL/min/1.73 m2 in 2007. Although the advisory has subsequently been diminished and the ACR guidelines state that for the lowest risk gadolinium agents measuring kidney function is not obligatory, in my experience it has been difficult to obtain contrast in these patients. This meta-analysis of 16 studies with 4,931 patients with Stage 4 or 5 CKD did not find any with NSF. One cannot say that the risk is zero because of the sample size and indeed there have been some reports of NSF with these group II agents. Therefore the study supports a maximum risk of 0.07%, but it may be lower than that. The authors do acknowledge weaknesses of their study, including lack of blinding, lack of standards for diagnosing NSF, and the possibility that NSF might be diagnosed at an institution other than where the MRI was performed and therefore not documented. Also, the studies were heavily weighted to the agent gadobenate dimeglumine (Multihance; Bracco, Monroe Township, NJ) in 64% of patients. This agent has hepatobiliary secretion, which may be protective. The bottom line is that the risk of NSF is low with the group II agents and we should not harm our patients by denying them MRI contrast when it would otherwise help in their care. -Mark L. Moster, MD I agree with your bottom line Mark, and I am going to keep this study in my files next time a radiologist gives me push back about administering a group II gadolinium-based agent in a patient with CKD. This review article is reminiscent of the article we reviewed for the June 2018 issue of Literature Commentary (1). It is always nice to revisit some of the things we discussed in the past. Just so we are all on the same page, Table 2 is the ACR manual classification of gadolinium-based agents and risk of NSF (2). Let me add that I think it is important for all of us to check with our neuroradiologist(s), to confirm what group of GBCAs are being used. Here at Mayo, it is exclusively group II agents. Moster and Bhatti: J Neuro-Ophthalmol 2020; 40: 434-441 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Literature Commentary TABLE 2. The ACR manual classification of gadolinium-based agents and risk of NSF GBCA Group Gadodiamide (Omniscan; GE Health care, Chicago, IL) I Gadoversetamide (OptiMARK; Guerbet, Villepinte, France) I Gadopentetate dimeglumine (Magnevist; Bayer Health care, Leverkusen, Germany) Gadobenate dimeglumine (MultiHance; Bracco) I II Gadobutrol (Gadovist/Gadavist; Bayer Health care) II Gadoteridol (Prohance; Brazzo) II Gadoteric acid (Dotarem; Guerbet: Clariscan; GE Health care) II Gadoxetate disodium (Primovist/Eovist; Bayer Health care) III Risk of NSF Associated with the greatest number of NSF cases Associated with the greatest number of NSF cases Associated with the greatest number of NSF cases Associated with few, if any, unconfounded cases of NSF Associated with few, if any, unconfounded cases of NSF Associated with few, if any, unconfounded cases of NSF Associated with few, if any, unconfounded cases of NSF Data remain limited regarding a NSF risk, but for which few, if any, unconfounded cases of NSF have been reported GBCA, gadolinium-based contrast agent; NSF, nephrogenic systemic fibrosis. -M. Tariq Bhatti, MD 1. MartinDR, KalbB, MittalA, SalmanK, VedanthamS, MittalPK. No incidence of nephrogenic systemic fibrosis after gadobenate dimeglumine administration in patients undergoing dialysis or those with severe chronic kidney disease. Radiology. 2018;286:113-119. 2. ACR Committee on Drugs and Contrast Media. ACR Manual on Contrast Media. Reston, VA: American College of Radiology, Moster and Bhatti: J Neuro-Ophthalmol 2020; 40: 434-441 2020. Available at: https://www.acr.org/-/media/ACR/Files/ Clinical-Resources/Contrast_Media.pdf. You may recall that after we reviewed the Martin et al article in 2018, the authors withdrew the manuscript because the study was not conducted in full accordance with the IRB protocol, so I would question its credibility. That is why I chose to revisit the topic. -Mark L. Moster, MD 441 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited.