Forme fruste keratoconus (FFKC), recently termed very asymmetric ectasia with normal topography (VAE-NT), is a clinically significant condition characterized by normal topography and slit-lamp examination in one eye, whereas the fellow eye shows signs of keratoconus. This atypical manifestation implies an incomplete state of the disease, in which the cornea protrudes, causing irregular astigmatism and vision impairment (Henriquez et al., 2020). Early diagnosis of FFKC is crucial as it enables patients to proactively address the condition and prevent its progression into fully developed keratoconus (KC), thereby mitigating the risk of further vision loss (Rabinowitz, 1998).

The biomechanical properties of KC play a pivotal role in understanding FFKC. KC is a corneal disorder marked by alterations in the normal collagen fibril network, frequently exhibiting asymmetry between the two eyes. It has been proposed that the progression of keratoconus is driven by a biomechanical cycle of decompensation, involving corneal thinning, increased mechanical strain, and stress redistribution, initiated by a localized reduction in corneal material properties (Ruberti et al., 2011; Roberts and Dupps, 2014). In FFKC, subtle biomechanical abnormalities are often present despite the absence of overt clinical findings. These biomechanical alterations may precede morphological changes, potentially leading to progressive corneal protrusion, irregular astigmatism, and visual impairment if left undetected. To improve diagnostic precision, the term VAE-NT has been proposed to replace the previously used designation of FFKC (Ambrósio et al., 2023; Ambrósio et al., 2017; Hwang et al., 2018). Early identification of these subtle biomechanical alterations is critical for the diagnosis of VAE-NT, enabling timely intervention to prevent progression to clinically manifest keratoconus and preserve visual function.

Corneal biomechanics have gained significant attention in recent decades, with their importance recognized in several applications, including the measurement of intraocular pressure, evaluation of ectasia risk following refractive surgeries, and assessment of the effectiveness of corneal cross-linking (CXL) treatment (Roberts and Dupps, 2014; Herber et al., 2021; Ramm et al., 2019; Girard et al., 2015; Piñero and Alcón, 2015). Several in vivo methods have been developed to assess corneal biomechanics, among which the Corvis ST (Oculus, Wetzlar, Germany) is widely used. The Corvis ST uses an ultra-fast Scheimpflug camera that captures 140 frames over 32.11 ms, allowing detailed analysis of corneal deformation in response to an air puff stimulus (Tian et al., 2014). Analysis of the resulting deformation yields several dynamic corneal response (DCR) parameters, which correlate with corneal stiffness (Xian et al., 2023; Miao et al., 2023). These metrics include the stiffness parameter at first applanation (SP-A1) (Ambrósio et al., 2017; Zhang et al., 2021a), the deflection amplitude DA, and the ratio between the deflection amplitudes at apex and 2 mm away from apex (DA Ratio Max 2 mm) (Lu et al., 2022). Additional parameters, such as Ambrósio’s relational thickness to the horizontal profile (ARTh) (Tian et al., 2021a), the Corvis biomechanical index (CBI) (Wang et al., 2017), the stress–strain index (SSI) (Zhang et al., 2021b), and the Chinese CBI (cCBI) (Zhang et al., 2024), have all demonstrated clinical utility in the diagnosis of keratoconus (Ren et al., 2021).

The evaluation criteria for diagnostic indicators are primarily based on metrics derived from receiver operating characteristic (ROC) analysis, including the area under the ROC curve (AUC), sensitivity, specificity, and positive and negative likelihood ratios (Mandrekar, 2010). Although an ideal diagnostic indicator should exhibit high performance across all metrics, in practice, some indicators may not simultaneously achieve high AUC, sensitivity, and specificity. Therefore, clinicians often rely on their clinical experience to interpret these metrics, which can further complicate the diagnostic decision-making process. In addition, the intraclass correlation coefficient (ICC) is commonly used to assess the repeatability and stability of diagnostic indicators (Muller R, 1994). Therefore, this study aims to identify diagnostic indicators that demonstrate superior performance in terms of AUC, sensitivity, specificity, and ICC.

Multi-criteria decision analysis (MCDA) is a structured methodology developed to support decision-making processes involving multiple and often conflicting evaluation criteria (Talukder et al., 2018). As described by Keeney, MCDA provides a logical and systematic framework for evaluating options based on multiple criteria (Keeney, 1982). In this study, we used the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS), which is a widely used MCDA method. Hwang and Yoon (1981) originally proposed that TOPSIS evaluates alternatives based on their geometric proximity to an ideal solution and distance from a negative ideal solution. Due to its computational simplicity and robustness, TOPSIS has become one of the most widely adopted quantitative techniques in multi-criteria decision-making. Its applicability spans various domains and relies on a well-established mathematical foundation. It has been applied for more than three decades (Hwang and Yoon, 1981; Jahanshahloo et al., 2006), with extensive validation and documentation in the scientific literature (Huang et al., 2016; Yoon and Hwang, 1995). In the TOPSIS framework, the optimal alternative is defined as the one with the shortest distance to the positive ideal solution and the greatest distance from the negative ideal solution.

In this study, the validity, reliability, and predictive performance of Corvis ST DCR parameters for identifying VAE-NT were comprehensively evaluated using the TOPSIS approach. Based on sample size requirements and TOPSIS rankings, the seven highest-performing DCR parameters were selected and integrated to construct a novel composite biomechanical index for differentiating VAE-NT from normal corneas.

The steps followed in this study are illustrated in Figure 1.

In recent years, several composite diagnostic indices have been developed to diagnose keratoconus. Vinciguerra et al. (2016) introduced the CBI to differentiate between normal and keratoconus corneas. Ambrósio et al. (2017) developed the TBI, which exhibited superior diagnostic performance. Subsequent studies have consistently confirmed their high diagnostic accuracy for keratoconus; however, their performance in detecting FFKC has been comparatively limited (Tian et al., 2021b; Asroui et al., 2022; Wallace et al., 2023; Koh et al., 2019; Ferreira-Mendes et al., 2019; Kataria et al., 2019; Song et al., 2022). Although studies have shown that these indices can achieve an AUC value greater than 0.90, there remains a need to further improve sensitivity, specificity, likelihood ratios (positive and negative), and ICCs (Zhang et al., 2020). This situation requires clinicians to consider multiple diagnostic indices simultaneously when making clinical decisions, thereby increasing the complexity and burden of diagnosis. To address this issue, we developed a novel diagnostic index, VAE-NT. The VAE-NT index demonstrated robust diagnostic performance, achieving high AUC, sensitivity, specificity, favorable likelihood ratios (positive and negative), and ICC values in both the training and validation datasets. Furthermore, compared with the TBI, the VAE-NT index exhibited superior diagnostic efficiency and predictive performance in our dataset.

The primary distinction of the VAE-NT index from previous indices lies in its parameter selection methodology, which is based on the TOPSIS model. Unlike the CBI and TBI, which rely solely on AUC-based parameter selection, the VAE-NT index incorporates a TOPSIS-based multi-criteria evaluation approach (Ambrósio et al., 2017; Vinciguerra et al., 2016). In this study, the TOPSIS model was utilized to evaluate the Corvis ST parameters in terms of validity, predictive power, and reliability. Based on this evaluation, the parameters were ranked according to their TOPSIS-normalized scores. The top seven parameters, determined by the available sample size, were then used in a binary logistic regression analysis, and the model with the optimal diagnostic performance was retained. This study represents a novel integration of the MCDA technique and statistical modeling for the comprehensive evaluation of diagnostic parameters in keratoconus detection.

Corneal thickness is a well-recognized factor influencing the diagnosis of keratoconus. It can significantly affect the diagnostic performance of parameters used to differentiate VAE-NT or KC from normal corneas. It is important to note that corneal thickness was not matched during the development of the CBI and TBI (6, 34). In the present study, the PSM method was applied to balance the distribution of central corneal thickness and other potential confounders between the normal and VAE-NT groups (Kane et al., 2020). The matching process also ensured a consistent sample size across both groups. This approach improves the reliability of the subsequent ROC analysis by meeting sample size requirements and effectively controlling for the potential confounding effects of corneal thickness on diagnostic outcomes. In the validation dataset, which exhibited significant differences in central corneal thickness, the VAE-NT index maintained superior discriminatory ability between VAE-NT and normal eyes. Recently, Ambrósio et al. (2023) introduced the TBI version 2 (TBI_V2), which showed high diagnostic accuracy for detecting VAE-NT, with an AUC value of 0.945 (DeLong test, P < 0.0001), a sensitivity of 84.4%, and a specificity of 90.1%.

In our dataset, TBI_V2 demonstrated slightly lower diagnostic performance than the VAE-NT index, although it remains highly effective in identifying VAE-NT. Notably, the VAE-NT index is exclusively based on Corvis ST parameters, which reflect corneal biomechanical properties, whereas TBI_V2 integrates both biomechanical and tomographic data. These findings highlight the important contribution of biomechanical information to VAE-NT diagnosis, aligning with the well-established role of collagen disruption in keratoconus pathogenesis (Rabinowitz, 1998; Roberts and Dupps, 2014). The components of the VAE-NT index—including SP A1, SSI2, A1 Time, SP HC, DA Ratio Max (2 mm), DA Ratio Max (1 mm), and Integrated Radius—were selected based on their demonstrated effectiveness in distinguishing VAE-NT from normal corneas (Miao et al., 2023; Tian et al., 2021b; Huo et al., 2023; Zhang et al., 2022). Moreover, SP HC, DA Ratio Max (2 mm), DA Ratio Max (1 mm), and Integrated Radius have also been reported to differ significantly between normal and keratoconus corneas (Song et al., 2023).

This study has several limitations. First, the single-center design and relatively moderate sample size (n = 106) may limit the generalizability of the findings. Future studies involving larger, multi-center cohorts are warranted to validate the results. Second, the retrospective nature of data collection may introduce selection bias, highlighting the need for well-designed prospective studies. Third, although the VAE-NT index demonstrated promising diagnostic performance, its validity, reliability, and predictive values require further evaluation using independent external datasets from other clinical centers.

In this study, the VAE-NT index was developed to distinguish VAE-NT from normal eyes. It demonstrated high sensitivity, specificity, AUC, favorable likelihood ratios, and good reliability, indicating strong diagnostic potential. The use of the TOPSIS model enabled a comprehensive evaluation of diagnostic indicators, facilitating the selection of features with superior overall diagnostic strength and providing clinicians with a more objective decision-making reference. Further validation in larger, more diverse populations and with longer follow-ups is necessary to support clinical implementation.