. 2018 Oct 31;33(2):274–282. doi: 10.1038/s41433-018-0229-6

Table 1.

Diagnostic accuracy of OCTA

Author and date	Study design	Equipment/algorithm	Study sample	Key findings	Comment
De Carlo et al. (2015)	Observational, retrospective study to describe the characteristics of CNV on OCTA and to determine the sensitivity and specificity of OCTA in detecting CNV as compared with FFA. CNV of any type.	Angiovue (Optovue, Inc, Fremont, CA). SSADA (split-spectrum amplitude-decorrelation angiography) software algorithm with 70,000 A scans per second.	Study involved 72 eyes, 61 patients. Of these, 30 eyes of 24 patients underwent same-day FFA and OCTA as part of the diagnostic study.	CNV on FFA n = 8, CNV on OCTA n = 4, sensitivity = 50%, no CNV on FFA n = 22, no CNV on OCTA n = 20, specificity = 91%.	Very small sample size of n = 8 to determine sensitivity of OCTA vs. FFA. All CNV subtypes were included.
Gong et al. (2016)	Retrospective, observational study describing the morphological characteristics and efficacy of OCTA in detecting nAMD.	Angiovue (Optovue, Inc, Fremont, CA). SSADA (split-spectrum amplitude-decorrelation angiography) software algorithm with 70,000 A scans per second.	Eighty-six eyes of 53 patients. Twelve eyes were excluded due to poor-quality images, 8 eyes were excluded due to the absence of OCTA or FFA results.	nAMD on FFA n = 52, nAMD on OCTA n = 45, sensitivity = 86.5%, no nAMD on FFA n = 34, no nAMD on OCTA n = 23, specificity = 67.6%, OCTA false positives n = 11 and false negatives n = 7 PPV 80.4% and NPV 76.7%.	Images of poor-quality were excluded. Therefore the applicability of OCTA across a wider population cannot be assessed.
Inoue et al. (2016)	Multicentre, retrospective cohort study to determine the sensitivity and specificity of OCTA and OCT to detect type 1 CNV as compared with FFA. type 1 nCNVM, OCTA alone, FFA alone, OCTA and OCT vs. FFA and OCT as reference standard.	Angiovue (Optovue, Inc, Fremont, CA). SSADA (split-spectrum amplitude-decorrelation angiography) software algorithm with 70,000 A scans per second.	Study involved 105 eyes of 92 patients with type 1 CNV as confirmed on FFA and OCT. Ten eyes with other types of CNV excluded.	CNV on en-face OCTA and structural OCT n = 90, sensitivity 85.7%, CNV on FFA alone n = 70, sensitivity 66.7%, CNV on en-face OCTA alone n = 70, sensitivity 66.7%.	Not treatment-naive subjects. Unable to assess specificity. Only relates to type 1 CNV. En-face OCTA with structural OCT more sensitive than en-face OCTA alone.
Carnevali et al. (2016)	Study to determine the detection rate of treatment-naive quiescent choroidal neovascularization due to AMD by OCTA, compared with FFA/ICGA.	Two instruments used AngioPlex CIRRUS HD-OCT. Carl Zeiss Meditec Inc. Dublin CA. OMAG (optical microangiography) algorithm with 68,000 A scans per second. Angiovue (Optovue, Inc, Fremont California, CA). SSADA with 70,000 A scans per second.	Twenty-two eyes of 20 patients with treatment-naive CNV due to AMD. Twenty-two eyes of 22 patients with non-neovascular AMD acted as the control group.	Quiescent CNV on OCTA n = 18, sensitivity = 81.8%. No CNV on OCTA n = 22, specificity = 100%.	Small sample. Treatment-naive quiescent CNV patients instead of eyes with active nAMD. Control group with non-neovascular AMD.
Liang et al. (2016)	Observational, retrospective case series to characterise the features of CNV on OCTA and to evaluate whether OCTA can be used to identify clinical activity of CNV as compared with FFA.	Angiovue (Optovue, Inc, Fremont, CA). SSADA (split-spectrum amplitude-decorrelation angiography) software algorithm with 70,000 A scans per second.	Forty-five eyes from 35 patients with nAMD. Six eyes had same-day FFA and OCTA, used as sample for measurement of sensitivity and specificity.	Twenty-five of 45 eyes were found to be clinically active. Of these 25 eyes, 22 (88%) had a CNV on OCTA. Of six eyes with same-day FFA and OCTA, active nAMD on FFA n = 3, active nAMD on OCTA n = 2, sensitivity 66.7%. Non-active nAMD on FFA n = 3, non-active nAMD on OCTA n = 3, specificity 100%.	Sensitivity calculations based on a very small sample. OCTA features of CNV did not correlate with activity.
Faridi et al. (2017)	Prospective case series to determine the sensitivity and specificity of OCTA in diagnosing nAMD compared with FFA/OCT.	Angiovue (Optovue, Inc, Fremont, CA). SSADA (split-spectrum amplitude-decorrelation angiography) software algorithm with 70,000 A scans per second.	Seventy-two eyes of patients with treatment-naive nAMD, non-neovascular AMD or normal controls	nAMD on FFA n = 32. nAMD on en-face OCTA alone n = 26, sensitivity 81.3%. nAMD on en-face and cross-sectional OCTA n = 32, sensitivity 100%. No nAMD on FFA n = 40. No nAMD on en-face OCTA alone n = 37 (grader A) / n = 39 (grader B), specificity 92.5% (grader A)/97.5% (grader B). No AMD on en-face and cross-sectional OCTA n = 39 (grader A), n = 40 (grader B), specificity 97.5% (grader A)/100% (grader B).	Small sample. Images of poor-quality were excluded.