Rapid Point-of-Care Identification of Aspergillus Species in Microbial Keratitis

This diagnostic study examines the diagnostic performance of a lateral-flow assay in detecting Aspergillus species in patients presenting with bacterial or fungal infection.

Key Points

Question

Can microbial keratitis caused by Aspergillus species be diagnosed using a rapid lateral-flow device (LFD) from corneal scrape samples and a minimally invasive corneal swab sampling technique?

Findings

In this diagnostic study of 198 participants in South India from whom scrape and swab samples were collected, high sensitivity and high specificity were achieved with an LFD identifying Aspergillus species.

Meaning

Findings of this study suggest that a commercially available LFD along with a bespoke ratiometric analysis has high diagnostic accuracy and could be an alternative standard of care for the point-of-care diagnostics for microbial keratitis.

Abstract

Importance

Microbial keratitis (MK) is a common cause of unilateral visual impairment, blindness, and eye loss in low-income and middle-income countries. There is an urgent need to develop and implement rapid and simple point-of-care diagnostics for MK to increase the likelihood of good outcomes.

Objective

To evaluate the diagnostic performance of the Aspergillus-specific lateral-flow device (AspLFD) to identify Aspergillus species causing MK in corneal scrape and corneal swab samples of patients presenting with microbial keratitis.

Design, Setting, and Participants

This diagnostic study was conducted between May 2022 and January 2023 at the corneal clinic of Aravind Eye Hospital in Madurai, Tamil Nadu, India. All study participants were recruited during their first presentation to the clinic. Patients aged 15 years or older met the eligibility criteria if they were attending their first appointment, had a corneal ulcer that was suggestive of a bacterial or fungal infection, and were about to undergo diagnostic scrape and culture.

Main Outcomes and Measures

Sensitivity and specificity of the AspLFD with corneal samples collected from patients with MK. During routine diagnostic scraping, a minimally invasive corneal swab and an additional corneal scrape were collected and transferred to aliquots of sample buffer and analyzed by lateral-flow device (LFD) if the patient met the inclusion criteria. Photographs of devices were taken with a smartphone and analyzed using a ratiometric approach, which was developed for this study. The AspLFD results were compared with culture reports.

Results

The 198 participants who met the inclusion criteria had a mean (range) age of 51 (15-85) years and included 126 males (63.6%). Overall, 35 of 198 participants with corneal scrape (17.7%) and 17 of 40 participants with swab samples (42.5%) had positive culture results for Aspergillus species. Ratiometric analysis results for the scrape samples found that the AspLFD achieved high sensitivity (0.89; 95% CI, 0.74-0.95), high negative predictive value (0.97; 95% CI, 0.94-0.99), low negative likelihood ratio (0.12; 95% CI, 0.05-0.30), and an accuracy of 0.94 (95% CI, 0.90-0.97). Ratiometric analysis results for the swab samples showed that the AspLFD had high sensitivity (0.94; 95% CI, 0.73-1.00), high negative predictive value (0.95; 95% CI, 0.76-1.00), low negative likelihood ratio (0.07; 95% CI, 0.01-0.48), and an accuracy of 0.88 (95% CI, 0.73-0.96).

Conclusions and Relevance

Results of this diagnostic study suggest that AspLFD along with the ratiometric analysis of LFDs developed for this study has high diagnostic accuracy in identifying Aspergillus species from corneal scrapes and swabs. This technology is an important step toward the provision of point-of-care diagnostics for MK and could inform the clinical management strategy.

Introduction

There is an urgent and unmet need to improve outcomes for patients with microbial keratitis (MK) in low- and middle-income countries, where MK remains the second most common cause of unilateral visual impairment, blindness, and eye loss.^1,2,3 Disease severity and poor outcomes are skewed toward tropical and subtropical geographic regions, predominantly affecting working-age agricultural workers. The microbial etiology of MK across this population is approximately 50% bacterial and 50% fungal, with fungal keratitis (FK) being primarily caused by filamentous Aspergillus species or Fusarium species and associated with the worst outcomes.^2,4,5,6,7

Timely diagnosis of MK and early prescribing of appropriate antimicrobials are imperative to increase the likelihood of a good patient outcome.^6,8,9,10,11 However, bacterial and fungal MK cannot be distinguished by clinical presentation alone,¹² and the current diagnostic strategy has several limitations.^13,14 One limitation is the paucity of diagnostic tests suitable for use within community or primary health care settings.¹⁰ Lack of diagnostics in these settings is associated with reduced access to and engagement with the health care system, delays in presenting at tertiary care centers, increased economic burden to the patient (who typically has low socioeconomic status), and worse overall outcomes.^7,15,16,17

Diagnostic challenges also persist within tertiary care centers. While microbial culture (gold standard) and smear microscopy are commonly available, they require invasive sampling of the cornea (corneal scraping) for retrieval of microbiological material. Furthermore, culture may take 2 to 14 days to yield a diagnosis, which yields a positive result only approximately 50.5% of the time.^18,19 Smear microscopy yields a positive result only 27.3% to 79% of the time and is not able to identify the microorganism, which may be required for treatment strategies.¹⁸ In vivo confocal microscopy and polymerase chain reaction are not readily available and are often prohibitively expensive.²⁰ Additionally, these methods are inherently intractable to implement at a primary care setting due to the required skilled personnel for sample collection, processing, and interpretation as well as a laboratory infrastructure, as indicated by the wide range of reported success rates.^4,18,21,22

Alternative diagnostic strategies are urgently required for MK across the whole care pathway, particularly those that do not require specialists and a laboratory infrastructure or invasive corneal scraping for sample retrieval. Lateral-flow devices (LFDs) are rapid, simple to use, and appropriate to use within community and primary care settings as well as other clinical settings without a dedicated laboratory. These LFDs have demonstrated high accuracy in diagnosing a number of infectious diseases (including those caused by bacteria and fungi),^{23,24,25,26,27} particularly when coupled with emerging commercial digital readers or smartphone-based applications to objectively interpret device results.^28,29

To our knowledge, the use of LFDs has not been reported in the diagnosis of MK. We therefore conducted a diagnostic feasibility study of the commercially available Aspergillus-specific LFD (AspLFD; OLM Diagnostics) that is approved for the detection of Aspergillus species in cases of suspected aspergillosis.^30,31 As a primary objective, we evaluated the diagnostic performance of AspLFD to identify Aspergillus species causing MK in corneal scrape samples. As a secondary objective, we assessed the performance of AspLFD to detect Aspergillus species in corneal swab samples of patients with FK. We further compared the performance of AspLFD reporting methods, including visual inspection, commercial cube reader, and a ratiometric analysis that was developed for this study.

Methods

The Aravind Eye Hospital Institutional Review Board approved this diagnostic study, which adhered to the tenets of the Declaration of Helsinki.³² All participants provided verbal informed consent. We followed the Standards for Reporting of Diagnostic Accuracy (STARD) reporting guideline.

Study Design, Participants, and Sample Collection

This prospective diagnostic study was conducted at Aravind Eye Hospital, a tertiary eye hospital in Madurai, Tamil Nadu, India, between May 2022 and January 2023. All study participants were recruited at the corneal clinic at Aravind Eye Hospital during their first presentation. Patients aged 15 years or older were eligible for inclusion if they were attending their first appointment, were undergoing corneal scrape, and had clinical suspicion of FK and/or a fungal smear report with positive results (Gram stain or potassium hydroxide test). Swab samples were tested if the AspLFD scrape result was positive or weak positive by visual inspection. Patients without growth in culture results and no photographs of the AspLFD results were excluded from the analysis. Patients received no compensation or incentive for participating.

Corneal scrapes and swabs were collected from participants with MK at the same time as standard-of-care diagnostic sampling. All sample collection was conducted under topical anesthesia (0.5% proparacaine). Scrapes were collected by standard clinical procedure (1 each for Gram stain and 10% KOH [potassium hydroxide] wet mount, 1 for the culture, and 1 for the AspLFD analysis). Swabs were collected by rotating a mini-tip polyester swab (Puritan; Puritan Medical Products) on the corneal ulcer clockwise and counterclockwise 3 times. The scraping instrument or swab tip was immediately placed into a 100-μL sample buffer (OLM Diagnostics) in a 1.5-mL tube (Eppendorf Tube; Eppendorf SE). All sample collection was performed by a corneal fellow or consultant. Additional details on study materials and methods are provided in the eMethods in Supplement 1.

Statistical Analysis

Per the manufacturer instructions, 70 μL of the sample in buffer was loaded onto the AspLFD and inspected visually after 20 minutes by the laboratory microbiologist (R.G. and/or K.R.); visual scores were graded as positive, weak positive, or negative. An ancillary commercial cube reader that was calibrated for the AspLFD provided outputs as positive, very low positive, or negative. Most tests (83%) were conducted on the day of sample collection, and some (17%) were delayed by a median (IQR) of 1 day (1-2 days). All tests were included in the same analysis pipeline. A photograph of the LFD was taken using a smartphone for the bespoke ratiometric analysis, which was developed for this study with open-source software (ImageJ; Fiji). This ratiometric method provided binary positive or negative outputs (eMethods in Supplement 1).

Only AspLFD tests with corresponding culture-positive reports (gold-standard reference test) were included in the analysis. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and κ values were calculated with Prism 9.5.1 (GraphPad). Likelihood ratios (LRs) and overall accuracy were calculated with a diagnostic test evaluation calculator (MedCalc version 20.218; MedCalc Software Ltd). The results obtained by the AspLFD did not inform any clinical decisions or patient management.

Results

A total of 419 participants who were recruited into the study underwent corneal scrape testing (Figure 1), and 178 of these participants concurrently underwent corneal swab testing (Figure 2). Overall, 198 scrape samples and 40 swab samples were included in the AspLFD performance analysis. Of the 198 participants who met the inclusion criteria, 126 were males (63.6%) and 72 were females (36.4%), with a mean (range) age of 51 (15-85) years. A total of 89 (45%) and 64 (32%) of 198 study participants had taken antibiotic and/or antifungal treatment prior to presentation at the study site, with 23 of 35 patients (65.7%) with culture results positive for Aspergillus receiving antifungal treatment prior to presentation at the corneal clinic.

Figure 1. Standards for Reporting of Diagnostic Accuracy Flowchart of Patient Recruitment, Exclusion, and Inclusion for Corneal Scrape Analysis by Aspergillus-Specific Lateral-Flow Device (AspLFD) and Ratiometric Analysis.

Figure 2. Standards for Reporting of Diagnostic Accuracy Flowchart of Patient Recruitment, Exclusion, and Inclusion for Corneal Swab Analysis by Aspergillus-Specific Lateral-Flow Device (AspLFD) and Ratiometric Analysis.

The culture report for all 419 recruited participants is provided in eTable 1 in Supplement 1, and the scrape and swab culture results for those included in the AspLFD analysis are shown in eTable 2 in Supplement 1. Participant demographic characteristics and clinical baseline data are provided in eTable 3 in Supplement 1. Of the 198 scrapes and 40 swabs, the prevalence of Aspergillus species detected by the AspLFD was 17.7% (35 of 198) in the scrape samples and 42.5% (17 of 40) in the swab samples per the culture report.

Development of a Bespoke Analysis Tool

The interpretation of LFDs was inherently subjective, with positive, weakly positive, and negative outcomes possible. Determining whether weakly positive bands should be classified as a positive or negative test result plays a role in clinical uncertainty and can affect device performance and subsequent patient management. Of the 238 AspLFD tests analyzed in this study, 51 (21.4%) were scored as weak positive by visual inspection (eTables 5 and 8 in Supplement 1), and further assessment of device photographs did not show a clear delineation between weak-positive and positive scores in several cases (Figure 3). When comparing weak-positive AspLFD tests with culture reports, 8 of 51 (15.7%) had a positive result for Aspergillus species. To overcome the subjectivity of visual inspection, we deployed and assessed the performance of a commercial cube reader that was configured for use with AspLFD (105 scrape samples). Only 15 of 32 (46.9%) positive cube reader results had a positive Aspergillus species culture, resulting in a PPV of less than 0.5. An additional 10 samples were recorded by the cube reader as having a very low-positive result and included both culture-positive and culture-negative samples for Aspergillus species (eTable 4 in Supplement 1). This provided PPVs of only 0.38 (when very low positive was included as a positive reading) and 0.47 (when very low positive was considered a negative reading) (eTable 4 in Supplement 1).

Figure 3. Aspergillus-Specific Lateral-Flow Device (AspLFD) Scored by Visual Inspection.

Representative images of weak-positive (A) and positive (B) AspLFD visual scores are shown. A variety of test-line intensities were observed.

Due to this high level of misclassification, the cube reader was not a suitable LFD analysis tool in this context. To overcome the limitations of varying band intensities and operator subjectivity, we developed a simple method of quantification that uses a ratiometric approach to determine the intensity of the test band vs the control line (test band:control band [T:C];eFigure A in Supplement 1). A receiver operating characteristic curve plotting calculated ratios and culture reports demonstrated excellent sensitivity and specificity (area under the curve, 0.95) (eFigure in Supplement 1). The derivation cohort (n = 93) enabled a cutoff threshold of 0.1 or higher for a true-positive result to be selected (sensitivity, 0.90 [95% CI, 0.79-0.96]; specificity, 0.94 [95% CI, 0.89-0.96]; and LR, 14.01). This threshold was applied for each AspLFD result T:C value, allowing for binary positive or negative scoring and removing the ambiguous weak-positive category.

AspLFD Analysis of Corneal Scrapes

The AspLFD analysis of all patient scrape samples was scored by visual inspection and ratiometric approach (eTable 5 in Supplement 1). The sensitivity (0.91 [95% CI, 0.78-0.97] vs 0.89 [95% CI, 0.74-0.95]), NPV (0.98 [95% CI, 0.93-0.99] vs 0.97 [95% CI, 0.94-0.99]), and negative LR (0.11 [95% CI, 0.04-0.33] vs 0.12 [95% CI, 0.05-0.30]) of the device were equivalent between visual inspection and ratiometric analysis. However, the device specificity (0.95 [95% CI, 0.91-0.98] vs 0.77 [95% CI, 0.70-0.83]), PPV (0.79 [95% CI, 0.65-0.89] vs 0.46 [95% CI, 0.35-0.57]), positive LR (18.05 [95% CI, 9.09-35.84] vs 3.92 [95% CI, 2.92-5.27]), and accuracy (0.94 [95% CI, 0.90-0.97] vs 0.79 [95% CI, 0.73-0.85]) were all significantly improved (no overlapping 95% CIs) with ratiometric analysis compared with visual inspection (Table 1). The κ values comparing the LFD analysis to the culture report determined that visual inspection rendered a moderate agreement (0.489; 95% CI, 0.365-0.613), while ratiometric analysis had an almost perfect agreement (0.801; 95% CI, 0.692-0.909). Ratiometric analysis was most effective at reducing the false-positive scrape result from 39 (19.7%) to 8 (4.9%) while maintaining the proportion of true-positive results (32 [16.2%] vs 31 [15.7%]). Further details of AspLFD ratiometric false-positive and false-negative results are provided in eTable 6 in Supplement 1.

Table 1. Aspergillus-Specific Lateral-Flow Device (AspLFD) Analysis of Scrape Samples^a.

	Visual inspection (95% CI)	Ratiometric analysis (95% CI)
Sensitivity	0.91 (0.78-0.97)	0.89 (0.74-0.95)
Specificity	0.77 (0.70-0.83)	0.95 (0.91-0.98)
PPV	0.46 (0.35-0.57)	0.79 (0.65-0.89)
NPV	0.98 (0.93-0.99)	0.97 (0.94-0.99)
Positive LR	3.92 (2.92-5.27)	18.05 (9.09-35.84)
Negative LR	0.11 (0.04-0.33)	0.12 (0.05-0.30)
Accuracy	0.79 (0.73-0.85)	0.94 (0.90-0.97)

Abbreviations: LR, likelihood ratio; NPV, negative predictive value; PPV, positive predictive value.

^a N = 198 scrape samples. Disease prevalence = 0.18 (95% CI, 0.13-0.24).

AspLFD Analysis of Corneal Swabs

A major limitation of clinical MK diagnosis is the requirement for an invasive corneal scrape to retrieve microbiological material to be used in any in vitro diagnostic assay. An LFD is simple to use and could offer an elegant diagnostic solution within the community and regions where access to ophthalmologists and a laboratory infrastructure is limited; however, collection of biological material is still required. To address this requirement, we investigated whether minimally invasive corneal swabbing would collect enough material (antigen) to transfer to the AspLFD for accurate analysis.

Based on prespecified inclusion and exclusion criteria, 40 swab samples were included in the AspLFD analysis and distinctively had poor specificity (0.04 [95% CI, 0.00-0.21]) due to the high number of false-positive LFD results within this group, which were visually classified as weak positive. The corresponding AspLFD analysis for the scrape subgroup is shown in eTable 7 in Supplement 1.

The confusion matrix for swab-based AspLFD scoring is reported in eTable 8 in Supplement 1. The sensitivity (1.00 [95% CI, 0.82-1.00] vs 0.94 [95% CI, 0.73-1.00]), NPV (1.00 [95% CI, 0.65-1.00] vs 0.95 [95% CI, 0.76-1.00]), and negative LR (0 vs 0.07 [95% CI, 0.01-0.48]) were equivalent between visual inspection and ratiometric analysis. However, the device specificity (0.83 [95% CI, 0.63-0.93] vs 0.30 [95% CI, 0.16-0.51]), PPV (0.80 [95% CI, 0.58-0.92] vs 0.52 [95% CI, 0.35-0.68]), positive LR (5.41 [95% CI, 2.20-13.29] vs 1.44 [95% CI, 1.10-1.88]), and accuracy (0.88 [95% CI, 0.73-0.96] vs 0.60 [95% CI, 0.43-0.75]) were all greatly improved with ratiometric analysis compared with visual inspection (Table 2). In particular, the specificity increased from 0.30 (visual inspection) to 0.83 (ratiometric analysis) owing to the effective removal of false-positive results compared with visual inspection (from 16 [40.0%]) to 4 [10.0%]) while maintaining the proportion of true-positive results (17 [42.5%] vs 16 [40.0%]). The κ values showed that visual inspection rendered only a fair agreement (0.271; 95% CI, 0.082-0.460), while ratiometric analysis had a substantial agreement (0.750; 95% CI, 0.547-0.953) with cultures. Further details on the false-positive and false-negative AspLFD results are provided in eTable 9 in Supplement 1.

Table 2. Aspergillus-Specific Lateral-Flow Device (AspLFD) Analysis of Swab Samples^a.

	Visual inspection (95% CI)	Ratiometric analysis (95% CI)
Sensitivity	1.00 (0.82-1.00)	0.94 (0.73-1.00)
Specificity	0.30 (0.16-0.51)	0.83 (0.63-0.93)
PPV	0.52 (0.35- 0.68)	0.80 (0.58-0.92)
NPV	1.00 (0.65-1.00)	0.95 (0.76-1.00)
Positive LR	1.44 (1.10-1.88)	5.41 (2.20-13.29)
Negative LR	0 (0)	0.07 (0.01-0.48)
Accuracy	0.60 (0.43-0.75)	0.88 (0.73-0.96)

Abbreviations: LR, likelihood ratio; NPV, negative predictive value; PPV, positive predictive value.

^a N = 40 swab samples. Disease prevalence = 0.43 (95% CI, 0.27-0.59).

Discussion

Microbial keratitis is an ophthalmic emergency, and improving the diagnostic offering throughout the patient care pathway is essential to enable early and appropriate clinical management. The 2 most abundant fungal species across India are Fusarium species and Aspergillus species; however, compared with Fusarium species, Aspergillus species is associated with worse clinical outcomes and is reported to respond less well to natamycin, the first-line antifungal agent.^33,34,35 Therefore, rapid and early delineation between the 2 species is important in considering alternate treatment regimens for Aspergillus. Approaches to differentiate Aspergillus species from Fusarium species have considered filament diameter and branching angles when imaged by in vivo confocal microscopy, but these attempts have been unsuccessful,^36,37 and microbial culture remains the gold standard for differentiating the species. The AspLFD offers a clear advantage for this purpose and may enable the rapid implementation of alternative treatment strategies for Aspergillus species.

Culture positivity rates for MK corneal scrapings vary between centers but are typically low. In this study of the 419 patients recruited, only 62% (259) were culture positive. Often the poor sensitivity is attributed to low microbial yield within the collected scrape samples.^38,39 Compared with the culture-positive samples, the scrape analysis results demonstrated high sensitivity and NPV (0.89 and 0.97) and low negative LR (0.12), indicating that there were few false-negative results and that the samples tested on the LFDs contained antigen (an extracellular glycoprotein³¹) above the limit of detection. While not included within the formal AspLFD analysis, 5 of the 76 culture-negative, fungal smear–positive scrape samples also had positive AspLFD results, suggesting that these patients likely had a missed Aspergillus species infection. A total of 23 of 35 patients (65.7%) with a positive Aspergillus species culture reported using topical antifungal treatment prior to clinical presentation (although the duration and recency of use were not recorded) (eTable 3 in Supplement 1), which did not impede the diagnostic performance of the AspLFD test. This finding was in contrast to studies reporting the performance of AspLFD in cases of aspergillosis, wherein patients undergoing antifungal therapy at the time of sample collection demonstrated reduced device sensitivity compared with treatment-naive patients.^40,41,42

Similarly high sensitivity (0.94), high NPV (0.95), and low negative LR (0.07) were attained for the swab samples compared with the scrape samples, indicating that sufficient antigen was collected from the ocular surface by this minimally invasive collection method. This finding is of particular importance because swabs could be used to collect corneal samples from patients with suspected MK in settings where the routine method of specimen collection by scraping is not possible, such as primary care.

A major limitation identified in this study was the high level of subjectivity needed for visual inspection. Subjectivity played a role in a large proportion of AspLFD tests being classified as weak positive (51 of 238 [21.4%]) by visual inspection. Previous studies of AspLFD tests for diagnosing aspergillosis classified weak-positive bands as positive results, improving their sensitivity and specificity.^{42,43,44,45,46} This finding was not appropriate for the MK samples, with this classification being associated with poor PPVs (ranging from 0.39 to 0.52). The commercial cube reader, which was formatted for use with the AspLFD (aspergillosis indication), was also an inadequate tool for correctly classifying MK weak bands.

To overcome this limitation, we developed a simple method to semiquantify the LFD using open-source software and a smartphone to capture photographs of the devices. A number of approaches have been or are being developed to semiquantify or quantify LFDs, which include use of either additional hardware^47,48,49 or a smartphone.^50,51,52,53 For many of these approaches, image capture, including lighting, distance from the device, and focus, was standardized, and postprocessing image normalization and contrast enhancement were required. We developed this quantification method post hoc using the device images that were captured throughout the study. These pictures were not taken in a standardized manner, including image capture parameters or lighting conditions. To account for this lack of standardization with the analysis tool, we selected a ratio of test-to-control line as the analysis method to normalize the band intensity measurements of each test individually regardless of the image size, image quality, or lighting. This approach proved robust across all of the AspLFD tests evaluated, with similar sensitivity, specificity, PPV, and NPV reported across the scrape and swab samples (eTable 10 in Supplement 1).

The ratiometric analysis showed a reduction in the total number of false-positive samples identified by visual analysis from 55 of 238 (23.1%) to 12 of 238 (5.0%), whereas the number of true-positive results was maintained (49 [20.1%] vs 47 [19.7%]). This reduction played a role in the dramatically increased specificity, PPV, positive LR, and accuracy of scrape (Table 1) and swab (Table 2) samples. Many of the scrape and swab samples had false-positive (or weak-positive) results by visual inspection. This finding may be a reflection of low-level contamination from the ocular mycobiome in this patient population, for whom Aspergillus species has been reported as an abundant constituent.^54,55,56,57 However, the semiquantitative method was able to accurately classify these LFDs.

Limitations

There are several limitations to this study. Most study participants presented to the corneal clinic after some delay and therefore had a relatively advanced stage of ulcer. Determining how the AspLFD performs with less severely established corneal ulcers is an important next step and would likely need to be conducted within the primary care system. We also anticipated that only approximately 20% of patients with MK would have a positive Aspergillus culture result. Due to resource constraints and anticipated limit of detection challenges (particularly from swabs), we chose to enrich the population with positive results by targeting patients with a positive fungal smear, and we examined only swab samples from patients whose scrape sample had a positive AspLFD reading. This study found that limit of detection was not an issue, and therefore it would be suitable to avoid this sample enrichment in the future for a more global approach. Additionally, this study combined analysis of corneal samples that were tested by LFD on the day of collection with analysis of samples that were tested on subsequent days. The diagnostic accuracy appeared comparable between the 2 groups (eTable 11 in Supplement 1), but the sample size for the subsequent days’ analysis was small. Further studies are warranted to determine the association of sample storage and duration with LFD accuracy.

Conclusions

This diagnostic feasibility study has demonstrated that AspLFD is appropriate for the rapid detection of Aspergillus causing MK from scrape and swab samples. We believe that this study is an important step toward the provision of point-of-care diagnostics suitable for primary care and other health care settings without a laboratory infrastructure. However, further diagnostic innovation is warranted, which can aid in the initial triaging question of bacteria or fungi and/or in rapidly indicating drug resistance profiles to establish whether and which antibiotics or antifungals should be initially prescribed to these patients.

Accurate and early diagnosis of MK is essential to rationally initiate clinical management and improve patient outcomes. Point-of-care diagnostics have revolutionized disease management for a number of conditions but are lacking for MK. Lateral-flow devices are low cost and simple to use. The AspLFD used in this study, coupled with the simple bespoke ratiometric analysis, proved to be a robust, highly sensitive and specific tool for identifying Aspergillus species from scrape and swab samples. This diagnostic feasibility study suggests that the LFD technology could offer an alternative gold standard approach for the diagnosis of MK (and enable MK diagnosis at the primary care level for the first time), particularly if multiplexed LFDs are developed to simultaneously screen several important MK pathogens. This technology is an important step toward the provision of point-of-care diagnostics for MK and could inform the clinical management strategy.

Supplement 1.

eMethods.

eTable 1. Spectrum of Microbial Isolates Cultured From Patients With Corneal Ulcer Recruited Into the Study

eTable 2. Spectrum of Microbial Isolates Cultured From Patients With Corneal Ulcer Included in the AspLFD Analysis

eTable 3. Patient Demographics Included in Analysis

eTable 4. AspLFD Scrape Sample Analysis – Cube Reader Matched Samples

eTable 5. Visual Inspection and Ratiometric Scoring of Scrape AspLFDs, and Corresponding Culture Results Confusion Matrix

eTable 6. Details of False Positive and Negative AspLFD Devices by Ratio Analysis (Scrape Samples)

eTable 7. AspLFD Scrape Sample Analysis – Swab Matched Samples

eTable 8. Visual Inspection and Ratiometric Scoring of Swab AspLFDs, and Corresponding Culture Results Confusion Matrix

eTable 9. Details of False Positive and Negative AspLFD Devices by Ratio Analysis (Swab Samples)

eTable 10. AspLFD Scrape & Swab Sample Ratiometric Analysis

eTable11. AspLFD Diagnostic Performance With “Same Day” or “Not Same Day” Samples (Ratiometric Analysis)

eFigure. AspLFD Semi-Quantitative Scoring

Supplement 2.

Data Sharing Statement