Randomized, Comparative Study of Full- and Half-Dose Fluorescein Angiography

Abstract

Purpose:

This study determines whether fluorescein angiography (FA) with a 250-mg dose of fluorescein (half dose) is equal in quality to the standard 500-mg dose of fluorescein (full dose) when using digital ultra-widefield (UWF) technology.

Methods:

In a randomized, prospective study using a UWF imaging system, FAs performed with half dose were compared with angiograms performed with full dose. Imaging studies were reviewed by 4 reviewers based on 6 characteristics: dye transit, macrovasculature, macula detail, microvasculature, leakage, and overall quality. The scores for macrovasculature, macula detail, microvasculature, and overall quality were converted to a fuzzy rating score to confirm results.

Results:

Seventy-nine FAs from 67 patients were reviewed for this study, including 12 patients who had both half-dose and full-dose FAs. Of all the factors studied, only microvasculature received a significantly different score between full dose and half dose that was confirmed by the fuzzy rating scale (3.79 vs 3.53; P = .04). Among those eyes that received both full and half dose, there was no significant difference in any of the 6 factors.

Conclusions:

In a UWF imaging system, aside from looking at fine microvascular abnormalities, the 250-mg dose of fluorescein provided similar results to a 500-mg dose. The images were not significantly different in overall quality.

Introduction

Fluorescein angiography (FA) involves the injection of fluorescein dye into the systemic circulation, illuminating retinal vessels whose images are captured by a fundus camera or a scanning laser ophthalmoscope. Novotny and Alvis published the first FA fundus photographs in 1961. ¹ They captured images every 12 seconds with a Zeiss Fundus camera using fast film and solutions that developed photographs within 10 minutes to enhance contrast. ¹ These photographs used 5 mL of 5% fluorescein along with a 5-mL bolus of normal saline. ^1,2

The early FA techniques were limited by the equipment available. In the following 20 years, there were improvements in technology. Better filters improved the ability to detect fluorescence and increased contrast. Newer cameras took images faster, and film could be developed more quickly. ³ FA converted from film to digital format after 1987, when Edward Feldman published the first digitized images. ³ By the early 2000s, nearly all FA studies were digitally created and stored, and in 2008, Friberg et al demonstrated that ultra-widefield (UWF) imaging systems could obtain excellent FAs of the posterior and peripheral retina in patients with diabetes. ⁴

The dose used by Novotny and Alvis also changed. Wessing, in collaboration with Zeiss, published a reference guide in 1969 to establish a standard for FA use, recommending 8 to 10 mL at 10% fluorescein. ⁵ Gass used 5 mL of 10% fluorescein for images published in his Stereoscopic Atlas of Macular Diseases, ⁶ first published in 1970, and for his descriptions of a wide variety of retinal pathologies. ^6,7 Despite the changes involved in FA techniques, the standard fluorescein dose in the past few decades has remained relatively constant at 500 mg, either 5 mL of 10% or 2 mL of 25%. ^{8 -10}

The first purpose of this study is to determine whether an FA study with 250 mg of fluorescein (half dose) is equal in quality to one with 500 mg of fluorescein (full dose) when using UWF imaging. The secondary purpose is to determine whether the incidence of adverse events is influenced by a half dose vs a full dose of fluorescein. We hypothesized that our retina specialists would assess a full-dose and a half-dose FA study as having no significant difference in quality.

Methods

We recruited 79 patients from July 2018 to August 2019 and received written consent for their participation. These patients had various retinal pathologies (Table 1), and each required an FA as ordered by the treating ophthalmologist. One eye of each patient (the eye first imaged) was entered in the study. Demographic data and medical history were collected through medical-record review. The dose of fluorescein (AK-fluor; Akorn Pharmaceutical), either 2.5 mL or 5 mL of 10%, was assigned by a random number generator. Experienced members of our imaging department prepared the syringes and injected the dye through the antecubital vein using a 25-gauge butterfly needle over 5 seconds. If the needle could not be inserted into the antecubital vein, the photographer would insert the needle through a vein in the wrist or hand. Patients were masked to the dose given. If a patient consented to participating in this study, underwent no reaction to fluorescein, and required another FA during the recruitment period, the patient was given the alternate dose and the same eye was entered in this study. All reported adverse events were recorded.

Table 1.

Retinal Pathologies Studied.

Pathology (N = 79 eyes)	Full dose	Half dose	Total
Proliferative diabetic retinopathy	12	14	26
Nonproliferative diabetic retinopathy	11	6	17
Focal chorioretinitis	4	3	7
BRVO	2	3	5
CRVO	2	2	4
Sickle cell retinopathy	1	2	3
Retinopathy of prematurity	1	1	2
Choroiditis with vasculitis	1	1	2
Other	7	6	13

Abbreviations: BRVO, branch retinal artery occlusion; CRVO, central retinal vein occlusion.

All FA studies were conducted with the Optos California camera system (Optos Inc). These images were deidentified and stored in portable network graphic (PNG) format. We excluded eyes with media opacities and patients who experienced extravasation from the final assessment. Eyes that met the exclusion criteria for media opacity were unanimously decided on by the director of the imaging department and a retina specialist.

Four retina specialists were asked to score each of the FA studies based on 6 factors: dye transit (time from the first appearance of dye in arteries to complete filling of arteries and veins, generally within 10 seconds), macrovasculature (appearance and clarity of large retinal vessels), macula detail (clarity of macula including delineation of the foveal avascular zone), microvasculature (ability to see retinal capillaries in posterior pole and periphery), leakage (diffusion of dye from retinal blood vessels or vascular structures), and overall quality. The retina specialist reviewers were masked to the dose of fluorescein given for each FA study. The reviewers graded these characteristics using a 5-point Likert scale (linguistic variables: very good [VG], good [G], neither good nor poor [N], poor [P], and very poor [VP]). For dye transit, we allowed the reviewers to score “not applicable” if the early dye transit was not recorded on the FA. Some nonexudative eyes had no leakage and were graded as “none.”

We converted the Likert scale scores for macrovasculature, macula detail, microvasculature, and overall quality into a fuzzy rating scale score to better normalize data and confirm results. ^{11 -13} Because of incomplete data sets for dye transit and leakage, these 2 factors were not converted into a fuzzy rating scale.

Fuzzy Rating Scale

The Likert scale is inherently subjective. The scale between linguistic variables varies for each reviewer, making it too “fuzzy” to compare the reviewers’ scores for each eye. We used a fuzzy rating scale to better elucidate the true scores of the reviewers. The fuzzy rating scale can confirm a more accurate difference in quality for our study.

First, we generated a set of Mamdani fuzzy rules to convert each linguistic variable (VG, G, N, P, and VP) on the Likert scale to a fuzzy set. This step is known as fuzzification. ¹¹ To fuzzify the data, we required 2 inputs and 1 output for each Mamdani rule. The first input was the opinion as the score given by the reviewer, and the second was the validity as the coefficient of variation (CV) of all the reviewers’ scores for that eye’s factor. The fuzzification output was the true score. We defined the opinion and true scores as VG, G, N, P, and VP. We defined the validity, which was the mean of the CV of all scores combined for all eyes, as high or low, where high was defined as a CV of less than 0.25. We made 10 Mamdani fuzzy rules (Figure 1). We used these rules to convert the Likert scale score to the true score. For example, if a reviewer gave a score of G but the CV among all reviewer scores for that eye’s factor was high, then the true score for that rating was N.

Figure 1.

Mamdani fuzzy rules to transform Likert scale score to true score (T). CV indicates coefficient of variation; G, good; H, high; L, low; N, neither good nor poor; P, poor; S, score; VG, very good; VP, very poor.

We used triangular membership functions to transform our true score data into a fuzzy set (Figures 2 and 3). Each true score was assigned a membership curve, with end points and midpoints that defined a fuzzy set. For example, a true score of N would have the fuzzy set (1, 3, 5) (see Figures 2 and 3).

Figure 2.

Membership curves to transform true score to triangular fuzzy number sets.

Figure 3.

Fuzzy sets generated from true scores and membership curves via fuzzification. G indicates good; N, neither good nor poor; P, poor; VG, very good; VP, very poor.

Once the fuzzy sets were created for each score, we aggregated the scores of each factor for each eye to generate an aggregated triangular fuzzy number (Formula 1). ¹⁰ If 4 reviewers had the true scores of N, G, N, and VG for macrovasculature of an FA, then the fuzzy set scores were {(1,3,5), (3,4,5), (1,3,5), (3,5,5)}. The aggregated triangular fuzzy number would be the average of each number within the fuzzy set, (2, 3.75, 5).

Formula 1.

For aggregated fuzzy set, X, the aggregated triangular fuzzy number (a_X, b_X, c_X), is the average of each number within the fuzzy sets (a, b, c) developed from Mamdani rules and membership function curves, where a, b, and c are integers, and n is the number of reviewers.

The aggregated fuzzy number was then converted to a crisp value (Formula 2) ¹⁰ to determine the center value of the membership curve generated by the aggregated triangular fuzzy number. An aggregated triangular fuzzy number of (2, 3.75, 5) would be converted to a crisp value of 3.63. In contrast, the Likert scale score of N, N, G, VG would be 3.75. The fuzzy rating scale pushes the true score closer to neutral, thereby better normalizing data and confirming differences between the half-dose and full-dose characteristics.

Formula 2.

This formula calculates the crisp value, C, where C is a real number and a_X, b_X, and c_X are real numbers from the aggregated triangular fuzzy number (a_X, b_X, c_X).

Statistics

Each characteristic mean score across all eyes was compared using a 2-sided unpaired t test. Patients who underwent both full- and half-dose FA were studied using a paired t test for each factor mean score. Significance was determined by an α error of less than .05. Statistical analysis was performed using STATVIEW 5.0.1 software (SAS Institute) and SPSS software version 24.0 (IBM). Membership function curves were created by MATLAB (MATLAB, 2014, 8.3.0.220, R2014a, The MathWorks Inc).

Results

Seventy-nine patients were recruited for this study. Of these patients, 12 were excluded from the study, and 12 patients underwent both a full-dose and a half-dose FA for the same eye at different times. Patients who met the exclusion criteria (n = 12) were excluded because of media opacities (33%), extravasation (42%), or poor venous access (25%). The diagnoses of the patients entered in the study are listed in Table 1. Of the 67 patients (79 eyes) included in the study, the mean age of the patients (SD) was 59.6 years (14.6 years), and 76% and 67% had a medical history of hypertension and diabetes, respectively (Table 2). There were no significant baseline differences between patients receiving full-dose and half-dose fluorescein. For those who underwent both a full-dose FA and a half-dose FA, the mean time (SD) between the 2 FAs was 5.4 months (3.5 months). Of all patients who underwent an FA, 1 patient experienced nausea at full dose, and 2 patients experienced urticaria at half dose (Table 3).

Table 2.

Baseline Characteristics of the Patient Population.

	Full dose	Half dose	Total (%)	P
Mean age ± SD	58.7 ± 14.9	59.7 ± 15.9	79	.77
Mean logMAR visual acuity ± SD	0.48 ± 0.55	0.65 ± 0.76	79	.27
Sex
Male	22	18	40 (51)	.42
Female	17	22	39 (49)
Race
African American	33	34	67 (85)	.58
White	5	6	11 (14)
Hispanic	1	0	1 (1)
Diabetes
Yes	25	28	53 (67)	.75
No	14	12	26 (33)
Hypertension
Yes	28	32	60 (76)	.55
No	11	8	19 (24)
Hyperlipidemia
Yes	20	18	38 (48)	.74
No	19	22	41 (52)
Smoking
Yes	17	19	36 (46)	.90
No	22	21	43 (54)
Eye laterality
Right eye	20	21	41 (52)	.99
Left eye	19	19	38 (48)

Table 3.

Adverse Events.

Adverse events (N = 8)	Full dose	Half dose	Total
Nausea	1	0	1
Urticaria	0	2	2
Extravasation (excluded from analysis)	4	1	5

Each retina specialist graded the FAs on 6 characteristics. There was no significant difference between the gradings of the half- and full-dose FA studies for each reviewer and all reviewers combined for dye transit, macrovasculature, macula detail, leakage, and overall quality (P > .1) (Figure 4; Table 4). The microvasculature score was significantly higher for full dose than half dose for 1 reviewer and when all the assessments were combined (P = .03 and P = .01, respectively) (see Figure 4; Table 4). Using a fuzzy rating scale, we confirmed the statistically significant higher score for full dose in the microvasculature subcategory (P = .04) (Figure 5; Table 4). Likert scale scores for dye transit and leakage were not converted to a fuzzy rating scale because the reviewers inconsistently used the score of “not applicable” or “none” for the eyes studied, leading to incomplete data sets for these factors. For the 12 eyes that underwent both full- and half-dose FAs, there were no significant differences for any factors (P > .1) (Figures 6 and 7; Table 5).

Figure 4.

Comparison of mean Likert scale scores for each and all reviewers for each factor between all full-dose and half-dose fluorescein angiographies. *P < .05.

Table 4.

All Half-Dose and Full-Dose Likert Scale Scores and Scores Transformed by Fuzzy Rating Scale for Each Factor.

Likert scale scores				Fuzzy rating scale scores
Factors, unpaired (N = 79)	Mean half dose (SD)	Mean full dose (SD)	P	Mean half dose (SD)	Mean full dose (SD)	P
Dye transit	3.33 (1.38)	3.35 (1.28)	.77	–	–	–
Macrovasculature	4.01 (0.84)	4.12 (0.90)	.21	3.85 (0.50)	3.93 (0.49)	.46
Macula detail	3.79 (0.92)	3.96 (1.05)	.18	3.69 (0.59)	3.84 (0.52)	.22
Microvasculature	3.55 (1.11)	3.85 (0.92)	.01a	3.53 (0.56)	3.79 (0.55)	.04a
Leakage	3.98 (0.96)	4.02 (0.87)	.99	–	–	–
Overall quality	3.76 (1.00)	3.85 (0.85)	.41	3.68 (0.56)	3.81 (0.47)	.28

^aP < .05.

Figure 5.

Comparison of mean scores transformed by a fuzzy rating scale for all eyes. *P < .05.

Figure 6.

Comparison of mean scores for all reviewers for each factor in eyes that received both full dose and half dose of fluorescein.

Figure 7.

Comparison of mean scores transformed by a fuzzy rating scale for eyes that received both half and full doses of fluorescein.

Table 5.

Likert Scale Scores for All Eyes That Received Both Half Dose and Full Dose and Scores Transformed by Fuzzy Rating Scale for Each Factor.

Likert scale scores				Fuzzy rating scale scores
Factors, paired (N = 12)	Mean half dose (SD)	Mean full dose (SD)	P	Mean half dose (SD)	Mean full dose (SD)	P
Dye transit	3.27 (1.32)	3.17 (1.4)	.67	–	–	–
Macrovasculature	4.02 (0.89)	3.98 (0.99)	.87	3.88 (0.57)	3.70 (0.61)	.25
Macula detail	3.71 (1.03)	3.81 (1.06)	.51	3.57 (0.60)	3.66 (0.49)	.98
Microvasculature	3.46 (1.07)	3.62 (1.13)	.44	3.54 (0.53)	3.48 (0.60)	.77
Leakage	4.03 (0.88)	3.83 (1.05)	.50	–	–	–
Overall quality	3.69 (0.97)	3.64 (0.97)	.65	3.68 (0.58)	3.59 (0.42)	.52

Conclusions

The techniques involved in performing FA today have changed dramatically since the 1960s. The most significant improvement lies in the technology used for capturing FA images, which has enhanced the contrast, widened the field of view, and improved image quality.

Other than slightly better visualization of the microvasculature with the higher dose, we found little difference between a dose of 5 mL (500 mg) and 2.5 mL (250 mg) of 10% sodium fluoride (NaFl) for digital UWF FA (Figure 8). Few publications have evaluated image quality with various doses. Justice et al in the late 1970s published the last prominent study addressing today’s standard dose of 5 mL of 10% fluorescein (500 mg), comparing it with a higher dose of 3 mL of 25% fluorescein (750 mg). ¹⁴ They determined that the FAs that used a higher concentration of fluorescein provided higher-quality images due to the enhanced contrast. ¹⁴ However, they relied on flash, film, and chemistry that limited the quality and contrast and could be overcome only with an increased concentration of fluorescein.

Figure 8.

Widefield fluorescein angiograms of the right eye of 1 patient taken 4 months apart with half dose on the left and full dose on the right. The brightness of the image was manually adjusted by the photographer, and there is very little difference in the quality of the angiograms.

Nasrallah et al published one of the first articles to suggest using less fluorescein could achieve similar results on imaging through improved film and chemistry. ¹⁰ They had 4 observers review 500 FA studies that used 2 mL of 10% fluorescein and were captured by a Zeiss Fundus camera with flash, film, and chemicals that enhanced the contrast. ¹⁰ Although there was no direct control group, the reviewers believed they were able to identify and diagnose pathology sufficiently with the decreased dose. ¹⁰

Nasrallah and his research team also noted a shortened transit time and decreased incidence of adverse effects in the reduced-dose cohort. ¹⁰ However, no statistics were provided in their study to support their claim. In our study, we did not find a difference in dye transit time, and we did not observe a significant reduction in adverse events with decreased dose. Research into the relationship between the dose and the incidence of adverse effects has been inconclusive. Yannuzzi and colleagues ¹⁵ determined that a lower percentage of patients experienced mild adverse effects with a 5% concentration of NaFl compared with 10% NaFl. Of patients who previously experienced adverse effects to fluorescein dye, Xu et al found that reducing the dose decreased the incidence of adverse effects on reexposure. ¹⁶ However, the small sample size led to statistically insignificant results. ¹⁶ Kwiterovich et al found that decreasing the dose of fluorescein did not significantly reduce the incidence of repeated reactions to fluorescein. ¹⁷

In a 2008 study, Moosbrugger and Sheidow concluded that a 2-mL dose of 10% fluorescein concentration produced significantly inferior quality images compared with the 5-mL dose. ⁹ They used the Topcon 50IX and the MegaVision 5000 digital camera to capture the FA images. ⁹ Our images were captured with the Optos California camera, which uses scanning laser ophthalmoscopy and UWF imaging of the retina. The UWF view can obtain approximately 82% of the retina in 1 photograph, identifying pathology in the peripheral retina more easily. ¹⁸ However, the peripheral view can be at a lower resolution. ^18,19 These cameras can capture FA images at high contrast thanks to the laser and confocal aperture. ^18,19 Although autocontrast control is now available for Optos California camera systems, we did not have this function for our study. Future studies will need to be completed to determine the efficacy of autocontrast control for half-dose FA studies.

One of the limitations of our study lies in our interrater reliability. We attributed the lack of consistency between raters to differences in their training in assessing FA studies and the inherent subjectivity of grading the quality of photographs. Because each retina specialist in our study assessed 79 FA studies, our reviewers may have altered their internal grading systems with time and practice. Our Likert scale was a subjective method of evaluating FAs, and we used a fuzzy rating scale to help reduce subjectivity. Fuzzy logic reduces interrater variability by taking into account each of the reviewer scores for each category in comparison with all of the other reviewers. The score is then optimized and scaled to match the other reviewers, and a composite score is produced. The fuzzy logic optimization prevents 1 reviewer from becoming an outlier compared with the others, reducing the variability.

Another limitation of our study is that there might have been variability in the monitors used by our retina specialists to interpret the images. The FAs viewed in the clinic were displayed in the DICOM (Digital Imaging and Communications in Medicine) format, and reviewers studied our FA images in PNG, a lossless format easily viewed on personal monitors. However, each reviewer consistently used a single monitor to assess FAs, thereby reducing intrarater variability.

The enhanced quality of digital, UWF FA images in this study allowed for an almost equal interpretation of images at a 500-mg dose vs a 250-mg dose of NaFl. Our reviewers graded full-dose FAs slightly higher on average when studying the microvasculature. For most eyes, this slight reduction in quality of imaging of the microvasculature would not significantly alter the interpretation of the study; however, for some patients with diabetes, fine microvascular abnormalities such as macular telangiectasia, or other microvascular diseases, full-dose FA may be preferred. This study demonstrates that excellent FAs can be obtained in patients with reduced renal function, for which a lower dose of fluorescein is desirable. We believe that 250 mg of fluorescein can provide fully adequate monitoring of retinal disease progression. When comparing the FA studies that use a 250-mg dose with the studies that use a 500-mg dose of fluorescein, the images are not significantly different in overall quality.