Limits...
Multivariate Analysis of the Ocular Response Analyzer's Corneal Deformation Response Curve for Early Keratoconus Detection.

Galletti JD, Ruiseñor Vázquez PR, Fuentes Bonthoux F, Pförtner T, Galletti JG - J Ophthalmol (2015)

Bottom Line: A combination of three factors including several corneal descriptors did not show better diagnostic performance than a combination of conventional indices.Conclusion.Conventional biomechanical indices seem to already provide the best performance when appropriately considered.

View Article: PubMed Central - PubMed

Affiliation: ECOS (Clinical Ocular Studies) Laboratory, Pueyrredón 1716, 1119 Buenos Aires, Argentina.

ABSTRACT
Purpose. To thoroughly analyze corneal deformation responses curves obtained by Ocular Response Analyzer (ORA) testing in order to improve subclinical keratoconus detection. Methods. Observational case series of 87 control and 73 subclinical keratoconus eyes. Examination included corneal topography, tomography, and biomechanical testing with ORA. Factor analysis, logistic regression, and receiver operating characteristic curves were used to extract combinations of 45 corneal waveform descriptors. Main outcome measures were corneal-thickness-corrected corneal resistance factor (ccCRF), combinations of corneal descriptors, and their diagnostic performance. Results. Thirty-seven descriptors differed significantly in means between groups, and among them ccCRF afforded the highest individual diagnostic performance. Factor analysis identified first- and second-peak related descriptors as the most variable one. However, conventional biomechanical descriptors corneal resistance factor and hysteresis differed the most between control and keratoconic eyes. A combination of three factors including several corneal descriptors did not show better diagnostic performance than a combination of conventional indices. Conclusion. Multivariate analysis of ORA signals did not surpass simpler models in subclinical keratoconus detection, and there is considerable overlap between normal and ectatic eyes irrespective of the analysis model. Conventional biomechanical indices seem to already provide the best performance when appropriately considered.

No MeSH data available.


Related in: MedlinePlus

Comparative diagnostic capacity of corneal biomechanical indices. Receiver-operating characteristic curves are plotted for central-corneal-thickness-corrected corneal resistance factor (ccCRF), timein, a combined index that includes ccCRF (BiomechScore), and the combination of the three extracted factors (3FactorScore). See Methods for details on each index. Area under the curve is specified in parentheses for each index.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC4446487&req=5

fig1: Comparative diagnostic capacity of corneal biomechanical indices. Receiver-operating characteristic curves are plotted for central-corneal-thickness-corrected corneal resistance factor (ccCRF), timein, a combined index that includes ccCRF (BiomechScore), and the combination of the three extracted factors (3FactorScore). See Methods for details on each index. Area under the curve is specified in parentheses for each index.

Mentions: With respect to diagnostic capacity (Table 3 and Figure 1), ccCRF ranked highest amongst the individual descriptors, with 71.3% specificity, 86.7% sensitivity, and an area under the curve of 0.85 (95% CI 0.79–0.91). The previously described combination of ccCRF and CH-CRF (biomechanical score (BiomechScore)) showed 81.6% specificity and 76.7% sensitivity, with an area under the curve of 0.87 (95% CI 0.82–0.93). The specificity and sensitivity of the optimum cutoff for the 12 best individual biomechanical descriptors are summarized in Table 3. The specified cutoffs were used to calculate a dichotomous “normal” or “abnormal” value for each case, and then the number of abnormal descriptors was counted for each observation to yield a descriptor score (9DescScore). In order to reduce multicollinearity, only the better performing descriptor of the pairs concerning the same aspects of the waveform (ccCRF and CRF, h2 and h21, and p2area1 and p2area) was considered, which led to a total number of 9 descriptors in the score (CRF, h21, and p2area were excluded). The 9DescScore had 90.8% specificity and 74.0% sensitivity with an optimal cutoff of >4 and an area under the curve of 0.89 (95% CI 0.84–0.94). Forward stepwise logistic regression of these 9 individual dichotomous variables resulted in a similar error rate (area under the curve 0.90, 95% CI 0.85–0.94) by including only 4 descriptors (ccCRF, CH-CRF, h2, and dive 2) and a more balanced performance in the control and keratoconic groups: 85.1% specificity and 78.1% sensitivity.


Multivariate Analysis of the Ocular Response Analyzer's Corneal Deformation Response Curve for Early Keratoconus Detection.

Galletti JD, Ruiseñor Vázquez PR, Fuentes Bonthoux F, Pförtner T, Galletti JG - J Ophthalmol (2015)

Comparative diagnostic capacity of corneal biomechanical indices. Receiver-operating characteristic curves are plotted for central-corneal-thickness-corrected corneal resistance factor (ccCRF), timein, a combined index that includes ccCRF (BiomechScore), and the combination of the three extracted factors (3FactorScore). See Methods for details on each index. Area under the curve is specified in parentheses for each index.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC4446487&req=5

fig1: Comparative diagnostic capacity of corneal biomechanical indices. Receiver-operating characteristic curves are plotted for central-corneal-thickness-corrected corneal resistance factor (ccCRF), timein, a combined index that includes ccCRF (BiomechScore), and the combination of the three extracted factors (3FactorScore). See Methods for details on each index. Area under the curve is specified in parentheses for each index.
Mentions: With respect to diagnostic capacity (Table 3 and Figure 1), ccCRF ranked highest amongst the individual descriptors, with 71.3% specificity, 86.7% sensitivity, and an area under the curve of 0.85 (95% CI 0.79–0.91). The previously described combination of ccCRF and CH-CRF (biomechanical score (BiomechScore)) showed 81.6% specificity and 76.7% sensitivity, with an area under the curve of 0.87 (95% CI 0.82–0.93). The specificity and sensitivity of the optimum cutoff for the 12 best individual biomechanical descriptors are summarized in Table 3. The specified cutoffs were used to calculate a dichotomous “normal” or “abnormal” value for each case, and then the number of abnormal descriptors was counted for each observation to yield a descriptor score (9DescScore). In order to reduce multicollinearity, only the better performing descriptor of the pairs concerning the same aspects of the waveform (ccCRF and CRF, h2 and h21, and p2area1 and p2area) was considered, which led to a total number of 9 descriptors in the score (CRF, h21, and p2area were excluded). The 9DescScore had 90.8% specificity and 74.0% sensitivity with an optimal cutoff of >4 and an area under the curve of 0.89 (95% CI 0.84–0.94). Forward stepwise logistic regression of these 9 individual dichotomous variables resulted in a similar error rate (area under the curve 0.90, 95% CI 0.85–0.94) by including only 4 descriptors (ccCRF, CH-CRF, h2, and dive 2) and a more balanced performance in the control and keratoconic groups: 85.1% specificity and 78.1% sensitivity.

Bottom Line: A combination of three factors including several corneal descriptors did not show better diagnostic performance than a combination of conventional indices.Conclusion.Conventional biomechanical indices seem to already provide the best performance when appropriately considered.

View Article: PubMed Central - PubMed

Affiliation: ECOS (Clinical Ocular Studies) Laboratory, Pueyrredón 1716, 1119 Buenos Aires, Argentina.

ABSTRACT
Purpose. To thoroughly analyze corneal deformation responses curves obtained by Ocular Response Analyzer (ORA) testing in order to improve subclinical keratoconus detection. Methods. Observational case series of 87 control and 73 subclinical keratoconus eyes. Examination included corneal topography, tomography, and biomechanical testing with ORA. Factor analysis, logistic regression, and receiver operating characteristic curves were used to extract combinations of 45 corneal waveform descriptors. Main outcome measures were corneal-thickness-corrected corneal resistance factor (ccCRF), combinations of corneal descriptors, and their diagnostic performance. Results. Thirty-seven descriptors differed significantly in means between groups, and among them ccCRF afforded the highest individual diagnostic performance. Factor analysis identified first- and second-peak related descriptors as the most variable one. However, conventional biomechanical descriptors corneal resistance factor and hysteresis differed the most between control and keratoconic eyes. A combination of three factors including several corneal descriptors did not show better diagnostic performance than a combination of conventional indices. Conclusion. Multivariate analysis of ORA signals did not surpass simpler models in subclinical keratoconus detection, and there is considerable overlap between normal and ectatic eyes irrespective of the analysis model. Conventional biomechanical indices seem to already provide the best performance when appropriately considered.

No MeSH data available.


Related in: MedlinePlus