If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
We sought to assess a smartphone-based, gold nanoparticle–based colorimetric lateral flow immunoassay paper sensor for quantifying urine 8-hydroxy-2′-deoxyguanosine (8-OHdG) as a biomarker for diabetic retinopathy (DR) screening.
Paper strips incorporate gold nanoparticle–8-OHdG antibody conjugates that produce color changes that are proportional to urine 8-OHdG and that are discernible on a smartphone camera photograph. Paper strip accuracy, precision, and stability studies were performed with 8-OHdG solutions of varying concentrations. Urine was collected from 97 patients with diabetes who were receiving DR screening examinations, including 7-field fundus photographs. DR was graded by standard methods as either low risk (no or mild DR) or high risk (moderate or severe DR). Paper sensor assays were performed on urine samples from patients and 8-OHdG values were correlated with DR grades. The differences in 8-OHdG values between the low- and high-risk groups were analyzed for outliers to identify the threshold 8-OHdG value that would minimize false-negative results.
Lateral flow immunoassay paper strips quantitatively measure 8-OHdG and were found to be accurate, precise, and stable. Average urine 8-OHdG concentrations in study patients were 22 ± 10 ng/mg of creatinine in the low-risk group and 55 ± 11 ng/mg of creatinine in the high-risk group. Screening cutoff values of 8-OHdG >50 ng/mg of creatinine or urine creatinine >1.5 mg minimized screen failures, with 91% sensitivity and 81% specificity.
Urinary 8-OHdG is a useful biomarker to screen DR. Quantitative 8-OHdG detection with the lateral flow immunoassay paper sensor and smartphone camera demonstrates its potential in DR screening. NOTE: Publication of this article is sponsored by the American Ophthalmological Society.
Since 2000, the increase in the global rate of diabetes has been staggering: from an estimated 151 million cases in 2000
By 2045, diabetes is projected to affect 628.6 million people, for an estimated prevalence rate of 9.9% of the estimated 9.9 billion worldwide population.
Paralleling the increasing rate of diabetes is the growing problem of diabetic eye disease. Worldwide, estimates are that as many as one third of people with diabetes are affected by diabetes-related eye complications.
International Agency for Prevention of Blindness, International Council of Ophthalmology, World Council of Optometry Strengthening health systems to manage diabetic eye disease: integrated care for diabetes and eye health.
Data from the Global Burden of Disease Study, extending from 1990 to 2010, indicate that 3.7 million people worldwide are visually impaired because of diabetic retinopathy (DR) and another 800,000 people are blind because of the disease. During the 2 decades from 1990 to 2010, blindness and visual impairment from DR grew at an alarming rate: there was a 27% increase in blindness and a 65% increase in visual impairment.
A major challenge in treating DR is its asymptomatic progression or “clinical silence.”
Diabetic Retinopathy Study Group Report number 6. Design, methods, and baseline results. Report number 7. A modification of the Airlie House classification of diabetic retinopathy. Prepared by the Diabetic Retinopathy Study Group.
have firmly established the vital role of early DR detection and treatment in preserving useful vision. It is now certain that the earlier DR is detected and treated, the better are vision outcomes.
Both the American Academy of Ophthalmology and the American Diabetes Association have developed DR screening recommendations to foster standardized screening practices and, hopefully, improve early DR detection rates.
Reasons for failure to screen are many, but may include the cost of examinations, the discomfort of pupil dilation and the bright lights that are required for retinal examination, lack of access to eye care, and the time required.
Various approaches for improving DR screening rates have been investigated. One strategy is the use of teleophthalmology. The advent of digital fundus photography coupled with a nonmydriatic fundus camera provides a relatively cheap method of obtaining screening retinal photographs without the need of a skilled photographer. Retinal photography has evolved as a useful method of DR screening.
The sensitivity and specificity of single-field nonmydriatic monochromatic digital fundus photography with remote image interpretation for diabetic retinopathy screening: a comparison with ophthalmoscopy and standardized mydriatic color photography.
An early teleophthalmology DR screening program was established in 2000 by the Indian Health Service–Joslin Vision Network Teleophthalmology Program to increase access to annual DR screening among American Indians and Alaska Natives.
A major limitation of these programs has been the need for trained specialists to assess the images. Telemedicine screening programs have had some encouraging results, but have not resulted in widespread acceptance or use.
is another promising technology for DR screening. With this technology, large sets of retinal images “train” and then “test” a computer algorithm designed to detect more than minimal DR. A DR diagnostic system using this technology, called IDx-DR (IDx Technologies, Coralville, IA, USA), recently received U.S. Food and Drug Administration (FDA) approval for the detection of more than minimal DR. In the IDx-DR clinical trial, non-mydriatic fundus cameras were placed in primary care sites where camera operators received minimal training. The sensitivity and specificity of the IDx-DR system, which is now commercially available, was evaluated in 900 patients with diabetes in a primary care setting to detect more than minimal DR. Fundus images analyzed by artificial intelligence were compared to dilated fundus photographs obtained by trained ophthalmic photographers. The IDx-DR system exhibited a sensitivity of 87.2% and a specificity of 90.7%.
These images can be graded by artificial intelligence–based screening software and have demonstrated excellent sensitivity of 99.1% and moderate specificity of 80.4% in detecting vision-threatening DR.
Deep learning with artificial intelligence also shows promise as a means to detect the presence of visually significant retinal disease in 3-dimensional optical coherence tomography (OCT) images, with the resultant rate of referral meeting or exceeding that of experts.
While the results of teleophthalmology and artificial intelligence screening are encouraging, each requires fundus photography, which may be a limitation to widespread use.
DR Screening with a Nanotechnology-Based Colorimetric urine “Dip-Stick”
In an effort to simplify DR screening, to make it less burdensome to patients, and to improve compliance with screening, we developed a urine-based colorimetric nanotechnology-based paper sensor, integrated into a smartphone for at-home measurements, to detect urinary concentrations of potential DR biomarkers. Our work has focused on developing a paper sensor for the measurement of urinary 8-hydroxy-2′-deoxyguanosine (8-OHdG), which is known to be a sensitive marker for DR and diabetic nephropathy.
We report our preliminary results with the use of this device to detect urinary 8-OHdG in patients with diabetes referred for retinal examinations.
The rationale behind our choice of urinary 8-OHdG as the biomarker for DR is as follows. Studies of DR development unanimously indicate that the first metabolic abnormality of hyperglycemia is oxidative stress, which in turn instigates and promotes several underlying downstream events.
Therefore, oxidative stress markers in the biological fluids of diabetic patients could be candidate biomarkers of the earliest occurrence of DR. Hyperglycemia leads to excessive production of superoxide radicals by 2 major events that result in oxidative stress. First, increased glucose concentrations cause excessive generation of superoxide radicals, which are the byproducts of electron transport chain reaction in mitochondria during glucose metabolism.
Second, hyperglycemia leads to auto-oxidation of glucose, which disrupts the voltage gradient across the mitochondrial membrane, resulting in abnormal reduction of molecular oxygen and generation of superoxide radicals.
Therefore, the mitochondria are the origin as well as the point of immediate impact of oxidative stress. Moreover, retinal mitochondria are predicted to be the most affected because the retina has the highest oxygen uptake and the highest glucose metabolism in the body.
In our preliminary studies we quantitatively evaluated urine 8-OHdG as a predictive biomarker of DR. As a urinary biomarker of DR, 8-OHdG offers a number of advantages: 1) levels of 8-OHdG directly correlate with oxidative stress,
Although 8-OHdG is an excellent biomarker for DR screening, it is not exclusively present in DR. Other conditions that may result in elevated urinary 8-OHdG levels include cancer, atherosclerosis, and diabetic nephropathy, as well as any condition that results in generalized cellular oxidative stress. Because urinary levels may be elevated in several conditions, 8-OHdG is appropriately considered as a screening biomarker rather than a diagnostic biomarker for DR. Elevated 8-OHdG levels signal the possibility of DR and the need for thorough fundus examination.
The use of 8-OHdG as a urinary biomarker for DR has been suggested by 2 clinical trials. One study included 60 patients and the other 96 patients, and both studies demonstrated that urinary 8-OHdG is more elevated in diabetic patients with DR than in those without DR,
suggesting that 8-OHdG may be a useful biomarker for DR.
We hypothesize that urinary 8-OHdG correlates with the presence of DR. Therefore, accurate quantification of urinary 8-OHdG can predict the onset of DR. Herein, we describe a nanotechnology-based, smartphone-integrated device that we developed and our preliminary results using this DR screening device to quantify urinary 8-OHdG in urine collected from 97 patients with varying degrees of DR.
A hallmark of this urine sensor is its ability to quantitatively and accurately quantify biomarkers in subnanomolar concentrations. By identifying biomarkers that demonstrate quantitative differences in diabetic patients with and without DR, this urine sensor can reliably detect transitions from minimal to more advanced stages of DR. Because the urine sensor is simple to use at home, and can be produced inexpensively, it has the potential to help patients comply with periodic DR screening and to seek an ophthalmic evaluation when the urine DR screening test suggests the presence of DR. This would markedly increase the potential for vision-saving early treatment for advancing DR as well as reduce the economic burden of yearly eye examinations.
Although quantitative assays such as enzyme-linked immunosorbent assay are readily available through laboratory evaluations, the urine sensor we have developed avoids the necessity of laboratory analysis while providing precise quantitative results.
Clinical Trial to Validate 8-OHdG as a Urinary Biomarker for Diabetic Retinopathy
To validate in patients with diabetes that urinary 8-OHdG is quantifiable with the paper sensor and can serve as a biomarker in DR screening, a prospective clinical trial was initiated. The study was approved by the Institutional Review Board of the University of Missouri, Columbia (IRB No. 2004053). Patients with type 1 or type 2 diabetes and who were >18 years of age were enrolled. Patients were excluded if they had received previous anti–vascular endothelial growth factor (VEGF) injection, photocoagulation for DR (grid or scatter) or retinal detachment repair, or if they had proliferative retinopathy or a history of cancer. Enrollment of patients was divided between those with no or minimal retinopathy and those with moderate or severe background retinopathy. Patients were not excluded if they had diabetic macular edema. Clinical examination was used to determine eligibility for enrollment. Color and red-free 7-field fundus photographs of each eye were obtained to allow subsequent DR classification. Urine samples were collected from each patient on the day of enrollment and were stored at −20 C until they were analyzed. If anti-VEGF injections or photocoagulation were required to treat DR, treatment was performed after fundus photographs and urine collection were completed. The 7-field fundus photographs were reviewed by a trained retina specialist (D.P.H.) and DR was classified according to the International Clinical Diabetic Retinopathy scale.
Design of Paper-Based Lateral Flow Immunoassay for 8-OHdG Detection
To validate 8-OHdG as a urinary biomarker, it was incorporated into the design of a simple sensor that could be used by patients at home to test their urinary 8-OHdG levels. We designed a gold nanoparticle (AuNP)–based lateral flow immunoassay for detection of urinary 8-OHdG. This technique uses competitive lateral flow immunoassay, AuNP-bound anti-8-OHdG antibody, and smartphone photography to determine 8-OHdG concentrations in urine. Integral to this sensor is the use of AuNPs, which can be bound to antibodies and are very stable.
A key characteristic of AuNPs is their reddish color caused by strong localized surface plasmon resonance phenomena. This color allows for the quantitative detection of the concentration of AuNP-bound anti-8-OHdG antibody by determining the grayscale density of a smartphone photograph of the bound AuNP on the paper sensor after exposure to urine. By binding anti–8-OHdG antibody to AuNPs, the grayscale density of AuNPs in the test strip is inversely correlated with the concentration of 8-OHdG in the urine sample. The greater the grayscale density of AuNPs at the test line, the lesser the concentration of 8-OHdG in the urine sample. The inverse relationship is based on the principle of competitive lateral flow immunoassay, which is more preferred than traditional sandwich methods in quantifying small molecules, such as 8-OHdG. Small molecules generally lack >1 antigenic determinant and therefore cannot be detected through double antibody sandwich assay. Inverse colorimetric assay has become standard clinical procedure for the identification of small molecule biomarkers.
Components of the Lateral Flow Immunoassay Paper Strips
The lateral flow immunoassay paper strips were fabricated by assembling 4 components: 1) the sample pad; 2) the conjugate pad; 3) the nitrocellulose membrane; and 4) the absorbent pad (Figure 1 and Appendix 1).
As shown in Figure 2, lateral flow immunoassay works by capillary force that drives the sample liquid to flow forward along the paper strip, thus encountering each of the 4 components. A drop of urine is introduced at the sample pad, and capillary force drives the sample to move along the strip. As the sample reaches the conjugate pad and dampens it, the AuNP–antibody conjugate is released from the pad into the urine specimen, and the 8-OHdG present in the sample binds to the antibody. It is important to note that some of the antibody bound to the AuNPs still remains free. The amount of free, or “unbound,” AuNP–antibody conjugate depends on the concentration of urinary 8-OHdG. The greater the concentration of 8-OHdG in the urine, the less the number of remaining unbound AuNP–antibody conjugates.
The whole mixture of urine and residual unbound AuNP–antibody as well as the bound complexes of AuNP–antibody–8-OHdG continue to move with capillary action along the paper strip to the test line on the nitrocellulose membrane. The residual free AuNP–antibody that is not bound by 8-OHdG in the sample then binds to the bovine serum albumin (BSA)–8-OHdG conjugate adherent on the test line and produces a red color corresponding to that of AuNP. As mentioned above, the color intensity at the test line depends on the concentration of unbound or free AuNP–antibody, which in turn is inversely proportional to the concentration of 8-OHdG in the urine. The residual sample then flows to the control line where antimouse immunoglobulin G binds to any residual AuNP–antibody complexes.
Using the Lateral Flow Immunoassay
To initiate the lateral flow assay, the “sample pad” end of the 3-mm wide strip was dipped into transparent plastic vials containing 80 μL of 8-OHdG standard solution or 80 μL of a previously collected urine sample. The development of color in the test and the control lines occurred within 10 minutes. The test and control lines were photographed with a smartphone camera. To maintain uniformity in the photography protocol across all the strips, the smartphone was kept fixed on a platform (of about 10 cm height) with its camera facing downward. To maximize the color intensity and to maintain uniformity, the photographs of the strips were taken 45 minutes after they were dipped into the standards/samples. After 45 minutes, the strips were slid under the camera platform, with the test and control lines facing the camera of the smartphone. The test and control line photographs were isolated according to protocol to obtain a central area of uniform density for each photograph. The intensity of the gray scale was quantified using ImageJ software.
Functionality Tests of Lateral Flow Immunoassay Paper Strips
Varying concentrations (0, 5, 10, 20, 40, 50, 60, and 70 ng/mL) of 80 μL of 8-OHdG solutions were added to a quartz cuvette and the paper strip was dipped into the solution and observed until color developed on the test line. Color development occurred within 10 minutes, and after 45 minutes the test line was photographed using a smartphone camera. The images were analyzed using ImageJ software and a standard curve was obtained by plotting the ratio of the mean gray values of test line/nitrocellulose membrane against the 8-OHdG concentration (ng/mL). Photographs were obtained with both an Android and an iPhone smartphone to compare the images of the lateral flow immunoassay AuNP density on the paper strip.
Interference studies were conducted to test whether other components of the urine interfere with 8-OHdG analysis. Commercially available artificial urine (Quik Fix) and 1× PBS were each independently used to make 8-OHdG standards (0, 5, 10, 15, and 20 ng/mL). Lateral flow immunoassay was done for both sets of standards and the results were compared.
To establish the confidence level associated with measurements made by the lateral flow immunoassay paper strips, uncertainty or error analyses were performed by conducting lateral flow immunoassays for standards of known 8-OHdG concentrations (30, 40, or 50 ng/mL) and analyzing the concentrations derived from lateral flow immunoassay standard curve. Each concentration was analyzed in duplicates.
To evaluate the precision of the lateral flow immunoassay paper strips, 18 different strips from the same batch were used to analyze and compare the 8-OHdG of 2 concentrations (5 and 30 ng/mL) measurements obtained with each.
Shelf Life Studies
To determine the shelf life of the lateral flow immunoassay strips, the paper strips made in a given batch were tested at 2 different time points; on the day of preparation and 2 months after the day of preparation (the strips were stored in a desiccator when not in use).
Lateral Flow Immunoassay of Urine Samples from Diabetic Patients
After determining the accuracy, precision, and stability of the lateral flow immunoassay paper strip device, we determined whether it could accurately screen for different stages of retinopathy by using the test to determine the 8-OHdG concentration in urine samples obtained from 97 patients with type 1 or type 2 diabetes enrolled in the prospective clinical trial. Urine samples were brought to room temperature and centrifuged at 3000 g to remove any solid material before testing. The supernatants were collected for further analysis. Creatinine values were analyzed using a commercially available creatinine assay kit by the University of Missouri Biorepository (CREP2l Cobas Roche, Mannheim, Germany). 8-OHdG was quantified using the lateral flow immunoassay paper strip. Because renal function can vary significantly, especially in patients with diabetes, 8-OHdG values were normalized with creatinine and expressed in terms of nanogram of 8-OHdG per 1 mg of creatinine.
Of the 97 patients enrolled in the study, 51 were males (52.6%) and 46 were females (47.4%). The average age was 61 years. Forty-one (42%) patients did not have apparent retinopathy, 16 (16%) had mild nonproliferative retinopathy, 25 (26%) had moderate nonproliferative retinopathy, and 13 (13%) had severe nonproliferative retinopathy. An additional 2 patients were found on fundus photography to have proliferative retinopathy that was not noted clinically at the time of study enrollment. Patients with no or mild retinopathy have a lower risk of progression to vision-threatening retinopathy when compared with patients with moderate or severe retinopathy.
Accordingly, patients were assigned to a low-risk group if they had no apparent or only mild nonproliferative retinopathy and to a high-risk group if they had moderate or severe nonproliferative retinopathy or proliferative retinopathy.
8-OHdG as a Urinary Biomarker for Diabetic Retinopathy
Dependability and Accuracy of Lateral Flow Immunoassay Paper Strips for Measurement of Urinary 8-OHdG
In the laboratory experiments, color development on the test line of the lateral flow immunoassay paper strip was found to occur with varying concentrations (0, 5, 10, 20, 40, 50, 60, and 70 ng/mL) of the 8-OHdG solutions (Figure 3). A standard curve of the ratio of the mean gray values of the test line/nitrocellulose membrane against the concentration (ng/mL) of 8-OHdG is shown in Figure 4. No statistical difference was found between the iPhone and Android smartphone cameras.
Fabrication of the lateral flow immunoassay strips, photography with the smartphone camera, and the analyses of the color intensity were all performed manually in this study. The lack of automation likely introduces subjective errors that may reduce the accuracy and precision of the device. Accuracy and functional standardization of the lateral flow immunoassay device will likely be enhanced with automated machine production and the use of a standardized cardboard phone positioning template for image acquisition and a smartphone “app” for standardized color strip analysis.
Lateral flow immunoassay was performed for standards made from artificial urine and 1× PBS and the results were compared. As shown in Figure 5, the color intensities of these 2 standards remain the same, indicating no interference from the urine matrix on the detection of 8-OHdG.
Error analysis of the lateral flow immunoassay paper strips was performed by comparing the calculated 8-OHdG concentration obtained with the paper strip to known 8-OHdG concentrations (30, 40, and 50 ng/mL). As shown in Table 1, the calculated concentrations obtained with the paper strip correlated well with those of known 8-OHdG concentrations. The average error was found to be as low as 2.75, validating the accuracy of the lateral flow immunoassay paper strips.
Table 1Accuracy of Lateral Flow Immunoassay Paper Strips for Measurement of 8-Hydroxy-2′-Deoxyguanosine: Error Analysis of Lateral Flow Immunoassay 8-Hydroxy-2′-Deoxyguanosine Values Compared with Known 8-Hydroxy-2′-Deoxyguanosine Concentrations in Solution
8-OHdG CC with LFIA
Absolute Error (CS − CC)
Percent Inaccuracy (Absolute Error/CS) × 100
Average error = 2.8
Average % inaccuracy = 2.8
8-OHdG = 8-hydroxy-2′-deoxyguanosine; CC = calculated concentration; CS = concentration of standards; LFIA = lateral flow immunoassay.
To evaluate the precision of the lateral flow immunoassay paper strips, different strips made from the same batch were used to analyze 8-OHdG samples of 5 and 30 ng/mL. The mean gray values remained the same, with the average percentage of deviation only 1.70% (5 ng/mL) and 0.59% (30 ng/mL; Table 2), clearly indicating high precision.
Table 2Precision Studies of Lateral Flow Immunoassay Paper Strips for Measurement of 8-Hydroxy-2′-Deoxyguanosine: Repeatability Analysis in 9 Test Samples of 2 Different Concentrations (5 ng/mL and 30 ng/mL)
The shelf life of the lateral flow immunoassay paper strips was evaluated by using strips from the same batch to perform 8-OHdG measurements on the same day the strips were made and again 2 months later. The mean gray values for 3 different test concentrations (10, 20, and 30 ng/mL) were almost identical to those obtained with freshly made or 2-month-old paper strips (Figure 6).
Urinary 8-OHdG Values in Low- and High-Risk Retinopathy
To compare the 8-OHdG levels obtained with the lateral flow immunoassay, we grouped patients into low- and high-risk groups as outlined previously. As Figure 7 shows, the results with lateral flow immunoassay paper strips exhibited a well-defined demarcation between low-risk DR and high-risk DR groups. The box plot distribution of normalized 8-OHdG values revealed no significant overlap. The average 8-OHdG value in the high-risk group was 55 ± 11 ng/mg of creatinine, which is 2.5 times greater than the average 8-OHdG value of 22 ± 10 ng/mg of creatinine in the low-risk group.
Figure 8, A shows the distribution of 8-OHdG values as measured by lateral flow immunoassay for each DR subgroup. The average 8-OHdG value in the no apparent retinopathy group is 21.9 ng/mg of creatinine; the mild nonproliferative retinopathy group, 44.4 ng/mg of creatinine; the moderate nonproliferative retinopathy group, 70.5 ng/mg of creatinine; the severe nonproliferative retinopathy group, 58.3 ng/mg of creatinine; and the 2 patients with proliferative retinopathy, 107.1 ng/mg of creatinine. As Figure 8, A shows, the average 8-OHdG values increase from no retinopathy to mild retinopathy to moderate retinopathy, but not from moderate retinopathy to severe retinopathy. If the outliers (defined as mean ± 2× standard deviation) in each of the groups are omitted, the separation between the no, mild, and moderate retinopathy groups increases (Figure 8, B). On analysis of the outliers, it was apparent that they are primarily caused by aberrant urinary creatinine levels. Urinary creatinine levels are a useful and necessary method of normalizing the data because of variations in renal function and potential 8-OHdG urinary excretion. However, at extremes of higher creatinine values, the normalized 8-OHdG values result in outliers that skew the data. A comparison of urinary creatinine values in the DR groups showed minimal correlation between urinary creatinine levels and DR, indicating that urinary creatinine values are not a useful independent biomarker of DR.
To minimize outliers, a urinary creatinine of >1.5 mg was used as a cutoff value. Figure 8, C shows the distribution of lateral flow immunoassay 8-OHdG values in the retinopathy groups when patients with urinary creatinine levels >1.5 mg were excluded. The average 8-OHdG value in the no apparent retinopathy group was 20.9 ng/mg of creatinine; the mild nonproliferative retinopathy group, 37.7 ng/mg of creatinine; the moderate nonproliferative retinopathy group, 73.5 ng/mg of creatinine; the severe nonproliferative retinopathy group, 60.6 ng/mg of creatinine; and the proliferative group remained the same at 107.1 ng/mg of creatinine. As Figure 8, C shows, the separation in 8-OHdG levels between the groups increases and many of the outliers drop out.
To determine the utility of the paper strip immunoassay of urinary 8-OHdG as a DR screening tool, a cutoff level is required that would indicate a high probability of vision-threatening DR. Accordingly, we assigned a cutoff urinary creatinine value of >1.5 mg, then grouped the remaining normalized 8-OHdG values into the 2 groups of low-risk DR (no apparent or mild nonproliferative retinopathy) and high-risk DR (moderate or severe nonproliferative retinopathy or proliferative retinopathy). Figure 9 shows the distribution of 8-OHdG values in the low-risk and high-risk groups after applying the urine creatinine threshold of 1.5 mg. With this cut off, the average 8-OHdG values were 21.2 ± 9.7 ng/mg of creatinine in the low-risk group compared with 61.1 ± 10.1 ng/mg of creatinine in the high-risk group. We examined our data to identify the optimal 8-OHdG cutoff value that would indicate an increased risk of vision-threatening DR and prompt an ophthalmic examination. A screening cutoff value of 50 ng of 8-OHdG/mg creatinine was found to result in four false-negative values, representing 3 patients with moderate nonproliferative retinopathy and 1 patient with severe nonproliferative retinopathy (Table 3). There were 8 false-positive values, representing 4 patients with no apparent retinopathy and 4 patients with mild nonproliferative retinopathy.
Table 3Diabetic Retinopathy Screening with Lateral Flow Immunoassay 8-Hydroxy-2′-Deoxyguanosine: Calculations of False Positives and False Negatives with Different Thresholds of 8-Hydroxy-2′-Deoxyguanosine Values and Creatinine Cutoff of 1.5 mg
Cutoff Value of Normalized 8-OHdG Plus Creatinine (>1.5 mg)
A sensitivity of 91%, specificity of 81%, negative predictive value of 0.9, and positive predictive value of 0.83 were calculated for a cutoff value of 50 ng of 8-OHdG. Values were determined within a range of 20 ng of the cutoff value (Table 4).
Table 4Diabetic Retinopathy Screening with Lateral Flow Immunoassay 8-Hydroxy-2′-Deoxyguanosine: Sensitivity, Specificity, and Positive and Negative Predictive Value Calculations with Different Thresholds of 8-Hydroxy-2′-Deoxyguanosine Values and Creatinine Cutoff of 1.5 mg
Cutoff Value for Normalized 8-OHdG Plus Creatinine (>1.5)
PPV calculated as TP/(TP+FP). NPV calculated as TN/(TN+FN). SP calculated as TN/(TN+FP). SN calculated as TP/(TP+FN). TP calculated as the total patients with DR − FP. TN calculated as the total patients with no DR–FN.
We also examined other factors that could potentially influence urinary 8-OHdG levels, such as age, gender, glucose control, and smoking (Figure 10, Figure 11, Figure 12, Figure 13). No correlation was found between the urinary 8-OHdG values and smoking, age, gender, or glucose control. An older age and female gender were associated with a more pronounced difference in 8-OHdG levels between the low- and the high-risk groups. Since this study was not longitudinal, the determination of glucose control was limited to the most recent hemoglobin A1c level (when available) or the patient's self-reporting of glucose control as “good” or “poor.” Concurrent hemoglobin A1c levels were not determined as part of this preliminary study. A longitudinal study would be required to determine any meaningful effect of glucose control on urinary 8-OHdG levels.
Potential Urine 8-OHdG Screening Strategy Based on the Study Findings
Thus, on the basis of the results in this preliminary study of 97 patients, a useful proposed DR screening strategy would require an initial screening ophthalmic examination in all patients with diabetes. If moderate or severe DR is diagnosed on examination, patients would be followed by their ophthalmologist as clinically indicated, just as is currently done. However, if no apparent retinopathy or if mild nonproliferative retinopathy is found, an annual lateral flow immunoassay urine screening, with a screen fail value of 50 ng 8-OHdG/mg creatinine or urine creatinine >1.5 mg, would be a reasonable follow-up strategy. A urine creatinine of >1.5 mg would qualify as a screen failure and subsequent ophthalmic examination would be required. An 8-OHdG ng/mg creatinine value of ≥50 ng would also constitute a screen failure and the patient would require a subsequent ophthalmic examination.
Such screening criteria applied to the study participants resulted in no false negatives, since the 4 patients with more advanced DR would have been identified by the initial clinical examination. Application of such screening criteria in our study population would result in 8 false positives: 4 with no apparent retinopathy and 4 with mild nonproliferative retinopathy.
The lateral flow immunoassay paper sensor device allows at-home use and requires only a smartphone camera to capture the color intensity of the test strip lines. Optimizing the physical location of the camera aperture to capture the test lines on the lateral flow immunoassay paper strip can be done with a simple folded cardboard positioning guide. A smartphone “app” can reliably localize the test strip on the photograph and compare the gray density to known standards, which allows patients to perform the test on a collected urine sample and determine their own 8-OHdG and creatinine levels. These levels are analyzed within the “app” and will result in a screening pass or fail based on the screening criteria outlined above. Thus, any patient with a smartphone and rudimentary knowledge of taking photographs with their smartphone can obtain reliable results.
The cost of producing a single lateral flow immunoassay paper strip with the reagents described is estimated to be <$1. The lateral flow immunoassay fulfills the criteria of a home-based test: it is simple to use and inexpensive to produce.
DR treatment is most effective when it is initiated early, but many diabetic patients do not receive regular DR screening to facilitate early diagnosis. The innovative lateral flow immunoassay for quantitative measurement of a urinary DR biomarker such as 8-OHdG has potential as a screening strategy that can improve DR screening practices, allowing for earlier diagnosis and treatment of DR. The availability of such a strategy would revolutionize the mode of DR screening by shifting away from yearly ophthalmic examinations to a simple “at-home” point-of-care device that can provide reliable and accurate quantitative measurements of DR biomarkers in the subnanomolar range. A challenge in refining this device for clinical practice will be determining the optimal single biomarker or multiple biomarkers and the cutoff values that minimize false negatives while accurately screening the majority of the population.
Although 8-OHdG is a good target biomarker for our assay, we did not observe a linear correlation between 8-OHdG concentrations and increasing severity of DR. The 8-OHdG/creatinine saturation level for our device is calculated to be in the range of 121-138 ng/mL, which may partially explain the nonlinear concentrations observed. The small number of patients within each group of moderate nonproliferative DR, severe nonproliferative DR, and proliferative DR may also partially explain these observations.
The principles of the lateral flow immunoassay can be applied to detect nearly any target molecule. Urine creatinine levels in the study participants were determined by a commercially available creatinine assay kit. However, the versatility of lateral flow immunoassay allows for a design that facilitates simultaneous analysis of creatinine and 8-OHdG levels on a single paper strip. This multiplexing ability of the lateral flow immunoassay allows for the placement of multiple biomarkers in parallel on the same paper strip. Our work to incorporate urine creatinine levels into the 8-OHdG test strip is in progress.
Several approaches may improve the accuracy of the lateral flow immunoassay device in DR screening. One strategy involves determining the levels of multiple biomarkers. Our data confirm that 8-OHdG is a good target biomarker. However, measurements of other biomarkers in addition to 8-OHdG may improve the ability to differentiate patients with no apparent DR or mild nonproliferative retinopathy from those with moderate or severe nonproliferative retinopathy. A reasonable additional target biomarker is angiopoietin-2. Angiopoietin-2 partially controls recruitment of pericytes and endothelial cell survival.
We are currently examining the angiopoietin-2 levels in the urine collected from our study patients, but do not yet have conclusive results.
In addition, in an effort to improve the accuracy of the lateral flow immunoassay screening device we have initiated a separate institutional review board–approved study of potential DR biomarkers in undiluted vitreous, blood, and urine samples. This prospective study includes 3 groups of patients undergoing vitrectomy surgery: 1) nondiabetic, (2) diabetic with no retinopathy, and 3) diabetic with moderate or severe nonproliferative DR or proliferative DR. Our aim is to use proteomic techniques to determine differences in proteins in the vitreous, blood, and urine collected from these 3 groups of patients. We postulate that the presence of ≥2 biomarkers will likely be most effective in DR screening of patients who are at greatest risk of vision-threatening retinopathy.
An additional concern for the use of the lateral flow immunoassay as a screening device is the need for an initial ophthalmic screening examination to exclude outliers. Such a requirement negates the lateral flow immunoassay as a “true” screening device. Instead, it would serve as a supplementary screening device after an initial ophthalmic examination. There may be instances in which patients at risk of vision-threatening DR would not be identified by lateral flow immunoassay screening and may not have an initial ophthalmic examination. One way to help mitigate this risk would be to require a prescription from health care providers to obtain the lateral flow immunoassay paper sensor test kit. In this way, the health care provider could counsel the patient to get an initial ophthalmic screening examination. However, given that an advantage of the lateral flow immunoassay device is the ability to use it when access to ophthalmic health care providers is limited, the requirement of an initial ophthalmic examination may cause some limitations of the usefulness of the device in these situations. Our research to identify additional complementary biomarkers may reduce the necessity of an initial ophthalmic examination.
We acknowledge the limited nature of the data from this preliminary trial of 97 patients. A much larger trial with longitudinal data is required to clarify the sensitivity and specificity of the biomarkers and appropriate cut off values. An important necessary milestone before initiating larger clinical trials is creating industrialized standards for the manufacture of the lateral flow immunoassay paper strips. Developing the capacity to create these paper strips in large quantities rather than manually in the laboratory will greatly enhance our ability to use them in widespread clinical trials.
Reducing barriers to DR screening is a key to the detection of vision-threatening disease at its earliest stage, when anti-VEGF therapy is most effective. The development of automated artificial intelligence–based DR screening of fundus photographs has the potential to improve the currently poor screening rates in patients. However, a limitation of artificial intelligence screening is the requirement of fundus photographs; either taken by a nonmydriatic camera at primary care sites or with a smartphone-based adapter that is clinically based. The advantages of a biomarker-based lateral flow immunoassay as we describe is the portability, low cost, and at-home use. The sensitivity (91%) and specificity (81%) of our biomarker-based system in this preliminary study is comparable with those of the artificial intelligence–based studies.
The quest to understand and treat DR began nearly 200 years ago, when Eduard von Jaeger reported the first case of DR, 4 years after Helmholtz's invention of the first ophthalmoscope (Appendix 2).
Until the development of argon laser photocoagulation in the 1970s, little could be done to treat DR. The availability of an effective treatment served as a motivator for incorporating DR screening into the standard of care of patients with diabetes. The first case report of intravitreal anti-VEGF injections for neovascular maculopathy occurred in 2005,
and anti-VEGF clinical trials have shown without a doubt that intravitreal anti-VEGF injections represent a major advance in the treatment of vision-threatening DR, particularly when intervention occurs in early-stage DR.
As our knowledge continues to expand on strategies for stopping DR progression, so too must our capabilities to detect DR at its earliest stage. The persistence of low compliance rates for DR screening underscores the need of innovative approaches to make periodic screening a part of each diabetic patient's self-management program. The device we have developed for the detection of urinary DR biomarkers, with its capacity for at-home monitoring, may offer a new direction for improving compliance with DR screening and preventing vision loss from this disease.
CRediT authorship contribution statement
Dean P. Hainsworth: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Supervision, Validation, Writing - original draft, Writing - review & editing. Abilash Gangula: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Writing - original draft, Writing - review & editing. Shreya Ghoshdastidar: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Writing - original draft, Writing - review & editing. Raghuraman Kannan: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Funding acquisition, Supervision, Validation, Writing - review & editing. Anandhi Upendran: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Writing - original draft, Writing - review & editing.
All authors have completed and submitted the ICMJE form for disclosure of potential conflicts of interest. Funding/Support: A portion of the support for this research came from the University of Missouri Coulter Translational Partnership Program. Financial Disclosures: Dr Hainsworth has equity ownership/stock options in Katalyst Surgical. The other authors indicate no financial support or financial conflict of interest. This manuscript is based on a thesis that was prepared in partial fulfillment of the requirements for membership in the American Ophthalmological Society and published in the Transactions of the American Ophthalmological Society in 2020. The manuscript underwent subsequent peer review by the Journal and has been modified following the peer review process. All authors attest that they meet the current ICMJE criteria for authorship.
The sensitivity and specificity of single-field nonmydriatic monochromatic digital fundus photography with remote image interpretation for diabetic retinopathy screening: a comparison with ophthalmoscopy and standardized mydriatic color photography.