The reliability of clinical methods in ophthalmology

Predicting Malignant Transformation of Choroidal Nevi Using Machine Learning

Objective

This study aims to assess a machine learning (ML) algorithm using multimodal imaging to accurately identify risk factors for uveal melanoma (UM) and aid in the diagnosis of melanocytic choroidal tumors.

Subjects and Methods

This study included 223 eyes from 221 patients with melanocytic choroidal lesions seen at the eye clinic of the University of Illinois at Chicago between 01/2010 and 07/2022. An ML algorithm was developed and trained on ultra-widefield fundus imaging and B-scan ultrasonography to detect risk factors of malignant transformation of choroidal lesions into UM. The risk factors were verified using all multimodal imaging available from the time of diagnosis. We also explore classification of lesions into UM and choroidal nevi using the ML algorithm.

Results

The ML algorithm assessed features of ultra-widefield fundus imaging and B-scan ultrasonography to determine the presence of the following risk factors for malignant transformation: lesion thickness, subretinal fluid, orange pigment, proximity to optic nerve, ultrasound hollowness, and drusen. The algorithm also provided classification of lesions into UM and choroidal nevi. A total of 115 patients with choroidal nevi and 108 patients with UM were included. The mean lesion thickness for choroidal nevi was 1.6 mm and for UM was 5.9 mm. Eleven ML models were implemented and achieved high accuracy, with an area under the curve of 0.982 for thickness prediction and 0.964 for subretinal fluid prediction. Sensitivity/specificity values ranged from 0.900/0.818 to 1.000/0.727 for different features. The ML algorithm demonstrated high accuracy in identifying risk factors and differentiating lesions based on the analyzed imaging data.

Conclusions

This study provides proof of concept that ML can accurately identify risk factors for malignant transformation in melanocytic choroidal tumors based on a single ultra-widefield fundus image or B-scan ultrasound at the time of initial presentation. By leveraging the efficiency and availability of ML, this study has the potential to provide a non-invasive tool that helps to prevent unnecessary treatment, improve our ability to predict malignant transformation, reduce the risk of metastasis, and potentially save patient lives.

Toluidine Blue 0.05% Vital Staining for the Diagnosis of Ocular Surface Squamous Neoplasia in Kenya

IMPORTANCE

Clinical features are unreliable for distinguishing ocular surface squamous neoplasia (OSSN) from benign conjunctival lesions.

OBJECTIVE

To evaluate the adverse effects, accuracy, and interobserver variation of toluidine blue 0.05% vital staining in distinguishing OSSN, confirmed by histopathology, from other conjunctival lesions.

DESIGN, SETTING, AND PARTICIPANTS

Cross-sectional study in Kenya from July 2012 through July 2014 of 419 adults with suspicious conjunctival lesions. Pregnant and breastfeeding women were excluded.

EXPOSURES

Comprehensive ophthalmic slitlamp examination was conducted. Vital staining with toluidine blue 0.05% aqueous solution was performed before surgery. Initial safety testing was conducted on large tumors scheduled for exenteration looking for corneal toxicity on histology before testing smaller tumors. We asked about pain or discomfort after staining and evaluated the cornea at the slitlamp for epithelial defects. Lesions were photographed before and after staining.

MAIN OUTCOMES AND MEASURES

Diagnosis was confirmed by histopathology. Six examiners assessed photographs from a subset of 100 consecutive participants for staining and made a diagnosis of OSSN vs non-OSSN. Staining was compared with histopathology to estimate sensitivity, specificity, and predictive values. Adverse effects were enumerated. Interobserver agreement was estimated using the κ statistic.

RESULTS

A total of 143 of 419 participants (34%) had OSSN by histopathology. The median age of all participants was 37 years (interquartile range, 32-45 years) and 278 (66%) were female. A total of 322 of the 419 participants had positive staining while 2 of 419 were equivocal. There was no histological evidence of corneal toxicity. Mild discomfort was reported by 88 (21%) and mild superficial punctate keratopathy seen in 7 (1.7%). For detecting OSSN, toluidine blue had a sensitivity of 92% (95% CI, 87%-96%), specificity of 31% (95% CI, 25%-36%), positive predictive value of 41% (95% CI, 35%-46%), and negative predictive value of 88% (95% CI, 80%-94%). Interobserver agreement was substantial for staining (κ = 0.76) and moderate for diagnosis (κ = 0.40).

CONCLUSIONS AND RELEVANCE

With the high sensitivity and low specificity for OSSN compared with histopathology among patients with conjunctival lesions, toluidine blue 0.05% vital staining is a good screening tool. However, it is not a good diagnostic tool owing to a high frequency of false-positives. The high negative predictive value suggests that a negative staining result indicates that OSSN is relatively unlikely.

Epidemiology of ocular surface squamous neoplasia in Africa

OBJECTIVES

To describe the epidemiology and an aetiological model of ocular surface squamous neoplasia (OSSN) in Africa.

METHODS

Systematic and non-systematic review methods were used. Incidence was obtained from the International Agency for Research on Cancer. We searched PubMed, EMBASE, Web of Science and the reference lists of articles retrieved. Meta-analyses were conducted using a fixed-effects model for HIV and cigarette smoking and random effects for human papilloma virus (HPV).

RESULTS

The incidence of OSSN is highest in the Southern Hemisphere (16° South), with the highest age-standardised rate (ASR) reported from Zimbabwe (3.4 and 3.0 cases/year/100 000 population for males and females, respectively). The mean ASR worldwide is 0.18 and 0.08 cases/year/100 000 among males and females, respectively. The risk increases with exposure to direct daylight (2-4 h, OR = 1.7, 95% CI: 1.2-2.4 and ≥5 h OR = 1.8, 95% CI: 1.1-3.1) and outdoor occupations (OR = 1.7, 95% CI: 1.1-2.6). Meta-analysis also shows a strong association with HIV (6 studies: OR = 6.17, 95% CI: 4.83-7.89) and HPV (7 studies: OR = 2.64, 95% CI: 1.27-5.49) but not cigarette smoking (2 studies: OR = 1.40, 95% CI: 0.94-2.09). The effect of atopy, xeroderma pigmentosa and vitamin A deficiency is unclear.

CONCLUSIONS

Africa has the highest incidence of OSSN in the world, where males and females are equally affected, unlike other continents where male disease predominates. African women probably have increased risk due to their higher prevalence of HIV and HPV infections. As the survival of HIV-infected people increases, and given no evidence that anti-retroviral therapy (ART) reduces the risk of OSSN, the incidence of OSSN may increase in coming years.

Interventions for squamous cell carcinoma of the conjunctiva in HIV-infected individuals

BACKGROUND

Squamous cell carcinoma of the conjunctiva is described in the ophthalmic literature as a rare, slow-growing tumour of the eye, normally affecting elderly men around 70 years of age. In Africa, however, the disease is different. The incidence is rising rapidly, affecting young persons (around 35 years of age), and usually affecting women. It is more aggressive, with a mean history of three months at presentation. This pattern is related to the co-existence of the HIV/AIDS pandemic, high HPV exposure, and solar radiation in the region. Various interventions exist, but despite therapy, there is a high recurrence rate (up to 43%) and poor cosmetic results in late disease. This review was conducted to evaluate the interventions for treatment of conjunctival squamous cell carcinoma in HIV-infected individuals.

OBJECTIVES

To evaluate the effect of interventions for treating squamous cell carcinoma of the conjunctiva in HIV-infected individuals on local control, recurrence, death, time to recurrence, and adverse events.

SEARCH METHODS

Using a sensitive search strategy, we attempted to identify all relevant trials, regardless of language or publication status, from the following electronic databases; PubMedPubMed, EMBASE and The Cochrane Library. We also searched clinical trial registries; WHO International Clinical Trials Registry Platform (ICTRP) and the US National Institutes of Health Clinicaltrials.gov. We searched the international conference proceedings of HIV/AIDS and AIDS-related cancers from the AIDS Education Global Education System (AEGIS). Searches were conducted between January and February 2012.

SELECTION CRITERIA

Randomised controlled trials (RCTs) involving HIV-infected individuals with ocular surface squamous neoplasia.

DATA COLLECTION AND ANALYSIS

We independently screened the results of the search to select potentially relevant studies and to retrieve the full articles. We independently applied the inclusion criteria to the potentially relevant studies. No studies were identified that fulfilled the selection criteria.

MAIN RESULTS

No RCTs of interventions currently used against conjunctival squamous cell carcinoma in HIV-infected individuals were identified.There is one ongoing RCT in Kenya that was registered in July 2012.

IMPLICATIONS FOR PRACTICE

Current clinical practice in treatment of squamous cell carcinoma of the conjunctiva rests on a weak evidence base of case series and case reports.

IMPLICATIONS FOR RESEARCH

Randomised controlled trials for treatment of this disease are needed in settings where it occurs most frequently. Preventive interventions also need to be identified. HIV/AIDS research has not focused on treatment of this tumour.

Measurement Validity in Ophthalmology

Measurements in ophthalmic research may be of parameters such as visual functions, quality of life or physical measures and can involve different types of instruments such as questionnaires and mechanical, chemical or electronic devices. Whatever the mode of measurement, however, all these devices require sufficient evidence for validity before inferences can be made on the basis of their findings. This article explores the nature and often overlooked importance of validity and explains some of the terminology involved. It discusses the main forms that ophthalmologists should be aware of before they can assess whether the instruments, old and new, are providing results upon which inferences can be made with any level of confidence. The literature search involved use of Medline, PubMed and Ovid as well as referenced papers in journals and books. Searches were comprehensive and international. Foreign texts were translated.

Langerhans cell histiocytosis

The clinical manifestations of Langerhans cell histiocytosis have been recognized for more than a century. For most of that time, physicians have viewed the disease from different perspectives, interpreting portions of its clinical spectrum as if they were distinct and unrelated entities. More recently, Langerhans cell histiocytosis has been unified into a single concept, though the disease continues to defy traditional classification. By most accounts, Langerhans cell histiocytosis appears to be a morphologically benign proliferation of inflammatory cells that escapes regulatory control mechanisms. Studies from patients with all stages of the disease, however, document clonal proliferation of immune processing cells (i.e., Langerhans cells), suggesting a malignant disease process. The most common ophthalmic manifestation of Langerhans cell histiocytosis is a solitary lesion of orbital bone, which typically responds to minimally invasive therapy. The best management of solitary orbital Langerhans cell histiocytosis is debatable and has been complicated by its recent designation as a risk factor for central nervous system disease. This article summarizes recent developments in understanding the biology of Langerhans cell histiocytosis, reviews its ophthalmic manifestations, prognosis, and the controversy surrounding treatment of isolated orbital disease.

Interventions for squamous cell carcinoma of the conjunctiva in HIV-infected individuals

BACKGROUND

Squamous cell carcinoma of the conjunctiva is a rare, slow-growing tumour of the eye, normally affecting elderly men around 70 years of age. In Africa, however, the disease is different. The incidence is rising rapidly, affecting young persons (around 35 years off age), and usually affecting women. It is more aggressive, with a mean history of three months at presentation. This pattern is related to the co-existence of the HIV/AIDS pandemic, high HPV exposure, and solar radiation in the region. Various interventions exist, but despite therapy, there is a high recurrence rate (up to 43%) and poor cosmetic results in late disease. This review was conducted to evaluate the interventions for treatment of conjunctival squamous cell carcinoma in HIV-infected individuals.

OBJECTIVES

To evaluate the effect of interventions for treating squamous cell carcinoma of the conjunctiva in HIV-infected individuals on local control, recurrence, death, time to recurrence, and adverse events.

SEARCH STRATEGY

Using a sensitive search strategy, we attempted to identify all relevant trials, regardless of language or publication status, from the following electronic databases; Medline/PubMed, CENTRAL, AIDSearch, EMBASE, LILACS, African Healthline, Cochrane HIV/AIDS Specialised Register, and the Cochrane Cancer Network Specialised Register. We searched the clinical trial register of the US National Institutes of Health, searched the international conference proceedings of AIDS and AIDS-related cancers, and contacted individual researchers, research organisations, and pharmaceutical companies that manufacture the drugs used as interventions. Searches were done between September 2005 and June 2006.

SELECTION CRITERIA

Randomised controlled trials (RCTs) involving HIV-infected individuals with ocular surface squamous neoplasia.

DATA COLLECTION AND ANALYSIS

We independently screened the results of the search to select potentially relevant studies and to retrieve the full articles. We independently applied the inclusion criteria to the potentially relevant studies. No studies were identified that fulfilled the selection criteria.

MAIN RESULTS

No RCTs of interventions currently used against conjunctival squamous cell carcinoma in HIV-infected individuals were identified.

IMPLICATIONS FOR PRACTICE

Current clinical practice in treatment of squamous cell carcinoma of the conjunctiva rests on a weak evidence base of case series and case reports.

IMPLICATIONS FOR RESEARCH

Randomised controlled trials for treatment of this disease are needed in settings where it occurs most frequently. Preventive interventions also need to be identified. HIV/AIDS research has not focused on treatment of this tumour.

Humphrey Matrix Frequency Doubling Technology Perimetry and Optical Coherence Tomography Measurement of the Retinal Nerve Fiber Layer Thickness in Both Normal and Ocular Hypertensive Subjects

PURPOSE

The purpose of this study was to determine by means of the Humphrey Matrix frequency doubling technology (FDT) perimetry and the optical coherence tomography (OCT) retinal nerve fiber layer (RNFL) thickness measurement whether functional and/or structural differences exist between normal and ocular hypertensive (OHT) subjects.

PATIENTS AND METHODS

One eye of 60 consecutive normal individuals and 60 OHT subjects was enrolled in this prospective observational comparative case series study. All subjects were examined at either the Ophthalmology Clinic, University of Chieti-Pescara, Chieti, Italy or the Department of Ophthalmology, S. Maria della Misericordia Hospital, Udine, Italy. All subjects underwent a full ophthalmic examination, including visual acuity, slit-lamp biomicroscopy, central corneal thickness ultrasound pachymetry measurement, achromatic automated perimetry, Matrix FDT perimetry, stereoscopic optic nerve head photography, and OCT. Matrix FDT perimetry mean deviation (MD), pattern standard deviation, glaucoma hemifield test, and 12 OCT RNFL thickness parameters were examined. Student t test, Bonferroni correction for multiple comparisons and receiver operator characteristics curve areas (AUROCs) were used to find any discrimination function between healthy and OHT eyes. Sensitivities at 83% and 92% specificities were reported.

RESULTS

The FDT MD scores ranged from -1.10 to +3.80 decibels (db) in normal individuals and from -4.75 to +3.20 db in OHT subjects. The comparison between the average MD in the 2 groups showed a statistically significant difference (P=0.024). OCT showed a statistically significant difference between the 2 groups when examining the ratio between the inferior and the superior mean RNFL thickness (P=0.004). For OCT, the parameter with the largest AUROC for discriminating between healthy and hypertensive eyes was the ratio between the mean inferior and superior RNFL thickness (AUROC=0.85, sensitivity=75% at specificity=83%, sensitivity=67% at specificity=92%). For Matrix FDT perimetry, the parameter with the largest AUROC was MD (AUROC=0.78, sensitivity=67% at specificity=83%, sensitivity=58% at specificity=92%).

CONCLUSIONS

Our results suggest that OHT eyes having a normal achromatic automated perimetry and a normal clinical disc appearance cannot be differentiated from normal eyes using conventional OCT parameters. When analyzing the ratio between inferior and superior average RNFL thickness, however, a significant difference is evident between healthy and hypertensive eyes. Using Matrix FDT perimetry, a significant difference in MD seems to exist between these 2 groups of eyes. The AUROCs confirm that OCT Iavg/Savg and Matrix FDT MD show the greatest sensitivity and specificity among the examined OCT and Matrix FDT parameters.

Statistical strategies to assess reliability in ophthalmology

Reliability of measurements and measurers is important so that we can trust the measurements we record. However, the statistical techniques used to assess reliability of measurements or measurers in the ophthalmic literature are often inappropriate, and not able to evaluate reliability between measurements/measurers. We review the techniques used in reliability studies for both continuous and categorical data, and describe appropriate statistical methods for particular study designs. We also highlight current techniques that are not appropriate in the analysis of reliability, but that are still commonly used in the ophthalmic literature. We hope that by highlighting these, we shall discourage their future use.

Digital surface microscopy analysis of conjunctival pigmented lesions: a preliminary study

The objective of this study was to investigate whether digital surface microscopy (DSM) could be used for the follow-up and comparison of malignant and benign conjunctival pigmented lesions (CPLs). Thirty-nine CPLs [16 de novo malignant melanomas (MMs), one MM arising from primary acquired melanosis (PAM), six PAMs and 16 naevi] were digitally analysed and biopsied. All of the PAMs and 10 naevi, which had not been surgically excised, were followed up using DSM. Thirty parameters were evaluated grouped into four categories: geometry, colour, texture and islands of colour. None of the CPLs that were followed up, which comprised 10 naevocytic naevi and seven PAMs, showed any morphological change at DSM analysis, except for one PAM which developed an MM 1 year later. Of the geometric variables examined, the area, maximum diameter and minimum diameter showed significantly higher values in MMs compared with benign CPLs. With regard to the colour of CPLs, MMs were significantly darker and bluer than naevi. In the texture group, contrast was significantly higher in MMs. In the islands-of-colour group, the imbalance of blue-grey regions and the presence of dark areas were significantly higher in MMs. DSM greatly simplified the follow-up of CPLs, such as PAMs with atypia, by providing satisfactory quality images with high reproducibility; this technique is also easy to use and well accepted by patients. Moreover, this preliminary study allowed us to determine which objective variables could be important for distinguishing between benign CPLs and conjunctival MMs.

Vertical cup-to-disc ratio: agreement between direct ophthalmoscopic estimation, fundus biomicroscopic estimation, and scanning laser ophthalmoscopic measurement

PURPOSE

Ophthalmoscopic estimation of the vertical cup-to-disc ratio (VCDR) of the optic nerve head is important in the management of patients with glaucoma or who are glaucoma suspects. The purpose of this study was to compare the accuracy of estimation of VCDR obtained with direct ophthalmoscopy with that obtained with fundus biomicroscopy through undilated pupils. Measurements of VCDR obtained with the Heidelberg Retina Tomograph (HRT), a confocal scanning laser ophthalmoscope, were used as a standard.

METHODS

Thirty young, healthy adults had their optic discs photographed and then imaged and analyzed with the HRT. Due to HRT software limitations, the VCDR could not be calculated automatically, and so a validated VCDR measurement was derived. This was used as the standard against which ophthalmoscopic estimations were compared. Two months later, the subjects had their VCDRs estimated using direct ophthalmoscopy and fundus biomicroscopy performed in random order after varying time intervals. Agreement between ophthalmoscopic VCDR estimation and HRT-derived VCDR measurement was assessed by means of bias plots and the weighted kappa statistic.

RESULTS

Compared with the HRT-derived VCDR measurement, both ophthalmoscopic techniques tended to underestimate VCDR. The bias with direct ophthalmoscopy was statistically significant. There were also wide variations in VCDR estimation with direct ophthalmoscopy and fundus biomicroscopy, which were more pronounced with direct ophthalmoscopy. The weighted kappa statistic indicated moderate agreement between fundus biomicroscopy and HRT-derived VCDR measurement. There was poor to fair agreement between direct ophthalmoscopy and HRT-derived VCDR measurement. The level of disagreement was independent of HRT-derived VCDR, optic nerve head size and pupil size for both direct ophthalmoscopy and fundus biomicroscopy.

CONCLUSIONS

Fundus biomicroscopy enables more accurate, less variable VCDR estimation than direct ophthalmoscopy. The clinician should record which method they used to examine the optic nerve head so that subsequent clinical decisions are not influenced by apparent VCDR changes.