The author «Araie M. Test-retest variability in structural par...»1 reports test-retest variability from 52 studies including planimetry from disc images, confocal scanning laser ophthalmoscope (HRT), scanning laser polarimeter (GDx), spectral domain optical coherence tomography (SD OCT), ultrasound biomicrocopy (UBM) and anterior segment optical coherence tomography (AS-OCT). In analyzing disease progression, either trend-type or event-type analysis is used. The confirmation of deterioration requires clear evidence which change from baseline exceeds the variability attributable to both the patient and the instrument, i.e. assessment of the test-retest variability is indispensable in determining the optimal frequency of performing imaging test. The test-retest variability of a system is estimated e.g. by calculating the coefficient of variation (CV), intraclass correlation coefficient (ICC), and minimum detectable changes (MDC). Coefficient of variation (CV) values < 10 % are considered capable of indicating good reproducibility. Intraclass correlation coefficient (ICC) 1 means perfect reproducibility (test-retest variablity = 0). ICC ≥ 0.9 is considered almost perfect and ICC ≥ 0.75 as a cut-off for good reproducibility. Minimum detectable change (MDC) can be calculated on the ICC and standard deviation (SD) of measurement results.
|Number of studies||CV min||CV max||ICC min||ICC max|
In HRT, test-retest variability has been extensively studied and reported to depend on patient age, severity of glaucoma, image quality, cylindrical error, lens opacity, surface geometry and reference plane. The rim area change in progressive eyes is reported to vary -0.005 - 0.012 mm2 per year. In order to detect 80 % of a yearly change of -0.012 mm2 with HRT, 16 examinations are needed over 4 years which produces 30 % false positives. If one takes 8 examinations over 4 years, 60 % of a yearly change of -0.012 mm2 will be detected with 20 % false positives.
In GDx test-retest variability is reported to be worse in more advance stages of glaucoma. 50 % of progressing eyes with 95 % specificity can be detected using the fast mode analysis of GDx Guided Progression Analysis (GPA).
Although the SD-OCT seems to provide better test-retest variability of the circumpapillary retinal nerve fiber layer (RNFL) and disc morphometric parameters, the systems need improvement in their test-retest variability measurement cababilities. Conservative estimate of MDC would be around 4 µm. When 50 µm (difference between normal and advanced glaucoma) is divided by 4 µm, glaucoma progression can be divide 13 stages, i.e. the same as with visual fields in the EMGT study (difference between normal and advaced glaucoma -25 db divided by the mean yearly change of 2 dB indiactinf progression).
Kommentti: Vaikka katsauksen kirjoittajan mukaan kliinikolle olisi hyödyllistä tietää parametrien toistettavuus, parametrien suurta lukumäärää ja niiden suuresti vaihtelevia toistettavuusarvoja on mahdotonta huomioida kliinisessä työssä.