Abstract
Purpose :
Laser flare photometry (LFP) provides prognostic information about patients with uveitis. The current standard approach for establishing final LFP values is to record the mean of 7 LFP measurements after removal of the highest and lowest measurements. We investigated alternative techniques for establishing a final value, in an attempt to identify a simpler approach to achieve the same standard.
Methods :
In a prospective study, LFP was performed on 200 eyes (100 patients with uveitis) using both Kowa FM-500 and FM-700 Laser Flare Photometers. A total of 7 measurements were obtained on each eye. We compared results between devices using 8 statistical methods: (1) all raw data; (2) mean of raw data; (3) mean after removal of highest and lowest values; (4) median of raw data; (5) median absolute deviation (MAD) outlier detection with mean (high threshold of 2.24); (6) MAD outlier detection with mean (low threshold of 1.28); (7) boxplot outlier detection (Carling method) with mean; and (8) boxplot outlier detection (Tukey method) with mean. To simulate fewer LFP measurements, analyses were repeated on the first 5 of 7 values. Intraclass correlation coefficient (ICC) was used to compare results. We also compared final values from different techniques on a per-eye basis. ICC of 0.75-0.90 were considered good.
Results :
Mean age of participants was 49.5±19.1 years; 72% were female. Using all 7 LFP measurements, ICC for comparison of the two devices did not vary substantially between techniques; ICC were as follows: (1) all raw data, 0.58; (2) mean of raw data, 0.80; (3) mean after removal of highest and lowest values, 0.81; (4) median of raw data, 0.81; (5) MAD outlier detection with mean (high threshold of 2.24), 0.77; (6) MAD outlier detection with mean (low threshold of 1.28), 0.77; (7) boxplot outlier detection (Carling method) with mean, 0.78; and (8) boxplot outlier detection (Tukey method) with mean, 0.78. ICC using first 5 measurements were similar to those using 7 measurements (range 0.58-0.81).
Conclusions :
All statistical methods demonstrated similar results between two devices and final values for alternative techniques were similar to those obtained from the current standard approach. Our study suggests that obtaining fewer LFP measurements and using simpler techniques for calculating final LFP values may be suitable for use in patient care and clinical research.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.