12.2  Investigation of Diagnostic Accuracy of OCT for Detection of DME

Fig. 12.1  Main Stratus OCT features in eyes with diabetic macular edema: cystoids changes (top), sponge-like thickening (middle), and edema associated with an adherent posterior hyaloid and a shallow serous retinal detachment (bottom)

12.2  Investigation of Diagnostic Accuracy

of OCT for Detection of DME

12.2.1  Potential Aims of OCT Testing in Clinical

Practice: Diagnostic Accuracy vs. Others

Medical tests are prescribed for several reasons, such as screening, diagnosis, monitoring course of the disease, selecting therapy or following effects of therapy, or determining drug levels or drug effects [9]. Many other factor have been recognized to influence the decision to use a test, including a number of patient’s or doctor’s characteristics and preferences, let alone the medical–legal value of the test results [10], often making the process of diagnosis subconscious [11].

Clinical research on the use of OCT in patients with DME is evolving together with the technological evolution of these devices, which allow recognizing an increasing number of retinal morphological details. Despite the fact that new information is provided, to be able to use it in practice means that OCT results should be proved to impact on patient outcomes, whether directly or through known predictors of prognosis and treatment effect, such as when the diagnosis of CSME is confirmed.

OCT is currently used in secondary and tertiary ophthalmic care services following biomicroscopic fundus

examination and is not used as a screening tool. As biomicroscopy is a part of every ophthalmic examination, it is not meant to replace FP/FB (such as for triage use), but rather to follow (i.e., using it as an add-on test) (see Bossuyt et al., for a methodological reference [12]). In practice, OCT can be used in patients in whom DME or CSME are suspected with the purpose of refining diagnosis, or can be used because of the properties not observed in FP/FB, such as recording retinal thickness to evaluate treatment response.

Among the potential roles of OCT when compared with those of FP/FB in patients with DR, we will discuss the following key issues:

1.Detection or confirmation of DME, or its severe form, CSME, such as when testing is used to rule in disease in people having signs of DR that may be associated with DME, who are referred by primary care professionals;

2.Exclusion of CSME, such as when testing is used to rule out disease in people referred to tertiary care services, e.g., when laser treatment has been suggested but the treating ophthalmologist believes that the presence of CSME is questionable (also medical–legal value in this case);

3.Establishing a baseline measure to monitor response in people who are treated for CSME to integrate imaging methods used in common practice (FP/FB and fluorescein angiography);

4.Other purposes, such as detecting or confirming the presence of a thickened adherent hyaloid, demonstrating chronic changes such a large intraretinal cysts, etc.

Items 1 and 2 refer to diagnostic accuracy, item 3 to monitoring of treatment response, and item 4 to presumed predictive role of some OCT findings in terms of prognosis and treatment response.

12.2.2  Diagnostic Accuracy of OCT for Detection of DME: Are Photography or Biomicroscopy a Valid Gold Standard?

Diagnostic accuracy refers to the ability to correctly diagnose disease, as defined using a gold standard test (or set of tests). The term “reference” standard is preferred to “gold” standard by methodologists because a perfectly valid test does not often exist. This is particularly true for DME which is traditionally defined by means of subjective assessment of retinal thickening using FP/FB. Being a subjective test, the interobserver agreement of FP/FB is substantial, yet far less than perfect, given that Cohen’s k was between 0.6 and 0.8 for most retinal lesions in the

ETDRS study [5], a fact that can limit the diagnostic performance of highly reliable OCT devices in a diagnostic test accuracy study. Furthermore, biomicroscopy does not allow recording of retinal thickening and is com-

12  monly used only to judge on its presence or absence. The advantage of OCT when compared with FP/FB in

diagnosing CSME is the ability to provide an objective, quantitative, highly reliable (see Appendix) measure of retinal thickness as well as additional morphological details (see last paragraph 2.4.).

In clinical practice, retinal-thickness measurement provided by OCT is always interpreted together with clinical examination, thus, integrating visual acuity and fundus biomicroscopy data. Using biomicroscopy, ophthalmologists can detect hard exudates in the macula which are a key component of the definition of CSME, although not the main finding using OCT. Therefore, the information provided by OCT assessment should always be interpreted together with fundus examination, as opposed to diagnostic test accuracy research, where reciprocal blinding of reference and index test results is a necessary quality item [9].

Summary for the Clinician

■Clinical examination using FP/FB is cheap and readily available, but also a subjective method. When biomicroscopy is used, there is no recording of the amount of thickening and thus, it cannot reliably monitor CSME except for a gross judgment on its presence or absence.

■Advantages of OCT when compared with FP/FB in patients with DME are its ability to obtain an objective and highly reliable quantitative measure of retinal thickness on a continuous scale, as well as to provide additional morphological details.

12.2.3  Diagnostic Accuracy of OCT to Detect CSME Using Time-Domain OCTs: How to Use OCT Retinal Thickness Cut-Offs?

In a meta-analysis including five studies, Virgili et al. [13] found that pooled sensitivity and specificity of OCT for detection of CSME were 0.79 (95% CI: 0.71–0.86) and 0.88 (95% CI: 0.80–0.93), respectively. The prevalence of CSME varied between 19 and 58% in the studies, thus being consistent with patients referred to secondary or tertiary care centers for a moderate to high suspicion of CSME. Two studies used an OCT 2000 and three

employed a Stratus OCT. The quality of these studies was often poor or unclear regarding a few methodological quality issues, which were assessed using the QUADAS checklist. Despite these limitations, the authors concluded that OCT devices are sufficiently accurate to diagnose CSME. Furthermore, by graphically inspecting metanalysis results, the authors found a possible threshold effect, i.e., that using values of 230–250 mm to define CSMEfavored sensitivity in three studies, while a cut-off of 300-mm-favored specificity in one study. Although this evidence on a threshold effect is sparse, these thickness cut-offs agree with values used to define CSME in studies by the Diabetic Retinopathy Clinical Research Network (http://public.drcr.net).

Central subfield thickness should be preferred to central point thickness [14] and was shown to be relatively unaffected by segmentation errors in fast-map measurements [15], provided that the standard deviation of the center point is <10% of central subfield thickness value [16].

In their review, Virgili et al. [13] used positive and negative likelihood ratios derived from their metanalysis to compute the range of usefulness of OCT to decide upon laser treatment of CSME, taking into account the benefits and making assumptions about adverse outcomes of photocoagulation. In fact, medical decisionmaking implies an a priori probability of disease which can be increased by positive test results or decreased by negative test results, taking into account the threshold for treatment, and readers can refer to classic evidence-based medicine books regarding this approach [17, 18].

For practical use, we suggest the following decision criteria. If we are rather suspicious, but not sure whether a patient is affected by CSME needing treatment, then finding a central subfield thickness of 300 mm or more with Stratus OCT is confirmatory, while values below 250 mm tend to exclude CSME. We suggest that values between 250 and 300 mm are less strong indicators of CSME presence. In this case, other findings should be considered to decide on laser photocoagulation, emphasizing, even more than usual, patient-reported vision, measured visual acuity, fluorescein angiographic findings, and fundus changes that one might consider to decrease the benefit of photocoagulation, such as an adherent and thickened hyaloid, subretinal fibrosis, or extensive macular ischemia (also see the following paragraph), despite the fact that research on these topics is limited. If the a priori suspect of CSME is very low or very high, then there is no point in obtaining an OCT to improve diagnostic accuracy because the test has no power to change such a strong prior belief. However, the test might be used for reasons other than CSME diagnosis, as described in the following paragraph.