Diagnostic accuracy of optical coherence tomography angiography for choroidal neovascularization: a systematic review and meta-analysis

Optical coherence tomography angiography (OCTA), based on optical coherence tomography (OCT), is an advanced imaging technology which has recently been approved for clinical ophthalmology application [1]. OCTA is considered as a noninvasive, rapid, secure and repeatable method, which is capable of displaying retinal vasculatures in three-dimensional assessment [2]. Without applying intraocular contrast agents, OCTA, a motion contrast imaging modality, depends on two sorts of backscattered light signals to visualize ophthalmic vascular networks in high-resolution. To be specific, one of the two signals is stable and derives from immobile structures such as neurosensory tissue, while the other is variable and dynamic over time generated by constantly moving tissues like erythrocytes. In comparison with stationary areas, erythrocytes can be measured by repeated B-scans. Thus, an image is created by detecting continuous changes of blood flow. This dye-less imaging technology renders confirmed diagnosis of ocular diseases without the restriction generated by contrast agents.

Pathological myopia, central serous chorioretinopathy (CSCR), age-related macular degeneration (AMD) and uveitides are commonly-detected diseases in ophthalmology, which are characterized by the formation of choroidal neovascularization (CNV) [3‐6]. In these disorders, neovascularization is commonly formed in the retina under specific stimulation [7, 8]. Neovascularization is highly permeable, in which hemorrhagic or exudation will occur with the progression of diseases. Under these conditions, visual impairment and irreversible blindness might present in the case of further exacerbated pathological changes.

The prevalence of ophthalmological diseases including pathological myopia and AMD is increasing year by year, which might lead to a potential risk of blindness in the population [9‐11]. Severe deterioration of visual acuity will occur by degrees if the progression of CNV is not prevented [12]. To reduce adverse consequences, early detection of CNV will be conducive to increase opportunities for timely and proper treatment. In addition, noninvasive imaging technology is also used for long-term monitoring on eyes which are at high risk of developing CNV [13]. In clinical practice, there are many diagnostic tools for ocular diseases including funduscopy, fluorescein angiography (FA), indocyanine green angiography (ICGA), conventional OCT and OCTA. As a new imaging method, OCTA shows advantages in detecting CNV [14]. In spite of the reported diagnostic accuracy of OCTA for CNV, results are inconsistent with no solid data confirming the exact accuracy of this method. To this end, this study was designed to evaluate the overall diagnostic value of OCTA in the detection of CNV by analyzing different studies.

Due to the advantages of OCTA in detecting microvascular pathology at various depths, previous studies had focused on the exploration of abnormal vascularity in the eyes by using OCTA images, however, these diagnostic performances failed to reach a consistent conclusion. In 2014, Moult E et al first reported the diagnostic value of OCTA in detecting CNV [17], revealing that 16 of 17 CNV eyes in exudative AMD patients were clearly visualized on OCTA images in comparison with 63 eyes from healthy individuals. The sensitivity and specificity of this exploration were 94 and 100%, respectively. Afterwards, De Carlo TE et al reported only four out of eight CNV eyes were clearly detected on OCTA, while all of the eight CNV eyes were definitely diagnosed in the article of Bonini Filho MA et al [19, 20]. Later, in 2016, Carnevali A et al in Italy and Gong J et al in China reported the diagnosis of CNV based on OCTA imaging [22, 25]. The sensitivity and specificity in the study of Carnevali A et al were 81.8 and 100%, respectively, while they were 86.5 and 67.6%, respectively, in the study of Gong J et al. Subsequently, Ahmed D et al, Faridi A et al and Nikolopoulou E et al also investigated the value of OCTA in detecting CNV in AMD [27‐29]. As a result, the sensitivity of these three studies ranged from 75.7 to 88% and the specificity varied from 90 to 100%. Herein, our meta-analysis analyzed the pooled efficacy of OCTA in detecting CNV, showing that the sensitivity, specificity and sROC-AUC were 0.87, 0.97 and 0.96, respectively. Therefore, OCTA technology might render rapid and safe exclusion of non-CNV eyes with high specificity for physicians.

In the elderly, AMD seems to be a major cause for visual impairment and the prevalence of AMD increases with age [33]. Due to the undefined cause of AMD, vascular endothelial growth factor (VEGF) inhibitors are widely applied in most treatments of AMD to prevent the growth of CNV [34]. Therefore, a useful method for monitoring the occurrence of CNV in AMD patients contributes to postponing the deterioration of this disease [35]. Currently, dye angiography is still used as a diagnostic golden standard for CNV; however, it is an invasive dye technology, with risks of side effects after injecting intravenous contrast agents [36]. In spite of accurate identification of CNV patients, the invasive injury could not be avoided for suspected patients without CNV. In addition, OCT is also a technology of detecting neovascular AMD, but it is not a single diagnostic tool owing to limitations in sensitivity and specificity [37]. Compared with OCT, OCTA is a more optimal technology and has been widely applied in the clinical recently. At present, AMD appears to be the most common disease, which might develop CNV. Therefore, in this study, seven articles were collected and analyzed, including 560 eyes that reported the diagnostic value of OCTA in only AMD related CNV. The sensitivity and specificity of OCTA to AMD related CNV were not significantly different from those of other ocular diseases (p = 0.12 and p = 0.44). In AMD patients, the occurrence of CNV in one eye may lead to CNV risk in the other eye. In a randomized clinical trial, 727 patients with CNV in one single eye were enrolled to receive a ranibizumab or bevacizumab therapy for the other eye. Two years after the treatment, 19% patients developed CNV in the other eye [38]. Thus, if OCTA technology could be performed to monitor CNV in the other eye, 88% of post-test probability in CNV positive eyes and 3% of post-test probability in CNV negative eyes would be reached. Together, these findings demonstrated that OCTA was a reliable imaging technology in early detection of CNV in AMD patients who had developed a CNV eye, which was also able to exclude individuals without CNV. For suspicious CNV eyes, it is unrealistic to receive constant dye angiography for monitoring disease progression, rather, it is promising to use OCTA as a surveillance and screening method.

In general, types of both study and device were potential sources of overall heterogeneity in this meta-analysis. The diagnostic value between the five prospective studies including 305 eyes was significantly different from that in the eight prospective studies including 473 eyes (p < 0.01). Compared with prospective studies, retrospective studies might underestimate diagnostic value due to the possible biases caused by inaccurate information recall by researchers. Moreover, to some extent, different kinds of devices might lead to imaging differences in diagnosis. As we have known, the projection artifact is a common problem affecting the diagnostic accuracy of the image equipment in detecting diseases. One of the reasons for projection artifacts is associated with the imaging equipment itself. The difference between AngioVue device group including 10 studies (426 eyes) is significantly different from other devices group including six studies (435 eyes) (P-value was 0.04). The sensitivity and specificity of AngioVue group were 0.90 and 0.96, respectively, while they were 0.83 and 0.98, respectively in other devices group. In addition, the reference standard was a source of heterogeneity in sensitivity between only FA group and other reference standards group (p < 0.01). The sensitivity in seven other reference standards studies including 312 eyes was higher than nine FA studies including 549 eyes (0.92 versus 0.83). Of note, FA is a common reference standard of the diagnosis of ocular vascular diseases due to the advantage of showing the leakage of neovascularization. The principle of OCTA is different from FA, for it is a quantitative equipment and provides neovascular network information. Souedan V et al used the multimodal imaging as the diagnostic reference standard to assess the diagnostic value of OCTA for CNV [30]. As a result, they found that OCTA showed a higher diagnostic value than FA, for the sensitivity and specificity of OCTA were 85.62 and 81.51%, respectively, and the sensitivity and specificity of FA for CNV were 74.5 and 82.35%, respectively. Besides, compared with FA, other clinical diagnostic methods such as ICGA and OCT also have their own advantages. Other reference standards including multimodal imaging may contribute to the correct classification of diseases, which may lead to a higher sensitivity in other reference standards group in our study.

There were certain limitations in our meta-analysis based on published articles. To begin with, in this study, we included all types of CNV to analyze the overall diagnostic value for all kinds of CNV. The diagnostic value of OCTA for different types of CNV might be different, however, the information of each study was insufficient to analyze the difference. Thus, we failed to analyze whether the type of CNV was the source of heterogeneity. Secondly, the sample size of several studies was less than 50 and some studies were retrospective ones in our meta-analysis. To this end, more prospective studies with large sample are warranted to confirm the diagnostic ability of OCTA for CNV in the future. Thirdly, only two studies reported the diagnostic value of OCTA only in Asian population, therefore, more studies from different countries are expected in the future.

In conclusion, OCTA is a potential and reliable method for detecting CNV in ocular diseases based on the sensitivity, specificity and AUC. It might be conducive to monitor the other eye in the case of CNV presence in one eye. In addition, more longitudinal studies of large sample are urgently needed to confirm this conclusion.