The purpose of the proposed cross-sectional study is to evaluate the effectiveness of some predictive models for axial length estimation, based on biometric data, through comparison with direct measurement made with the biometer itself. The study was conducted on a heterogeneous population of emmetropic and ametropic adult subjects. Exclusion criteria were established, including: positive history of trauma, surgery or current or past ocular disease that may affect biometry; children; subjects with unstable refraction; contact lens or orthokeratology lens wearers. The subjective refraction of the test population was determined by serving the phoropter while the objective refraction was obtained using the “i.Profiler” topographer and aberrometer. Biometric data collection was performed with the “Huvitz HBM-1” biometer, which provided measurements of axial length (AL) and corneal curvature (K1 and K2 with respective orientations). The data were transcribed into an Excel spreadsheet for processing. Consistent with the other studies analyzed, for each included sample, only the spherical equivalent was considered as the refractive reference. Three mathematical models were tested for the possibility of predicting axial length from corneal curvature and spherical equivalent, with subsequent comparison against direct measurements. Statistical analysis of the collected data allowed comparison of the estimated values with those obtained directly by biometer, assessing the degree of agreement and identifying any limitations of each model. For each subcategory considered, the R^2 and the average bias between calculated and measured values were determined. Limits of 95% agreement were also determined by a Bland-Altman analysis to assess the agreement between the two methods of obtaining AL data. the significance of the results was checked by calculating p-values. With reference to the total sample, the formula proposed by Kim et al. yielded biases ranging from +1.36 mm to +1.65 mm (p < 0.01), with wide limits of agreement and R2 ranging from 0.7905 to 0.9668 (p < 0.01), depending on the type of SE considered. In contrast, the Morgan et al. model produced bias between -0.11 mm and +0.05 mm (p > 0.05 for objective SE), with narrower limits of agreement than the previous model and R2 between 0.8263 and 0.8380 (p < 0.01). Finally, the Queirós et al. formula revealed bias in the range of +0.05 and +0.18 (p > 0.05 for subjective SE), with narrow limits of agreement and R2 between 0.8284 and 0.8436 (p < 0.01). In addition, further analyses were performed for different subgroups of subjects, which confirmed the heterogeneity of model applicability. As a result of the analysis, it was concluded that axial length prediction using mathematical models is not a substitute tool for proper ocular biometry, but rather a complementary one that can be applied only in clinical settings that are not highly rigorous.
Lo scopo dello studio trasversale proposto è quello di valutare l’efficacia di alcuni modelli predittivi per la stima della lunghezza assiale, basati su dati biometrici, attraverso il confronto con la misura diretta effettuata con il biometro stesso. Lo studio è stato condotto su una popolazione eterogenea di soggetti adulti emmetropi e ametropi. Sono stati stabiliti dei criteri di esclusione, tra cui: anamnesi positiva per trauma, chirurgia o patologie oculari in atto o pregresse che possano influenzare la biometria; bambini; soggetti con refrazione instabile; portatori di lenti a contatto o di lenti ortocheratologiche. La refrazione soggettiva della popolazione in esame è stata determinata servendosi del forottero mentre quella oggettiva si è ottenuta tramite il topografo e aberrometro “i.Profiler”. La raccolta dei dati biometrici è stata eseguita con il biometro “Huvitz HBM-1”, che ha fornito misurazioni di lunghezza assiale (AL) e curvatura corneale (K1 e K2 con rispettivi orientamenti). I dati sono stati trascritti in un foglio di calcolo Excel per l’elaborazione. Coerentemente con gli altri studi analizzati, per ogni campione incluso, è stato considerato come riferimento refrattivo il solo sfero equivalente. Sono stati testati tre modelli matematici per verificare la possibilità di predire la lunghezza assiale a partire dalla curvatura corneale e dall’equivalente sferico, con successivo confronto rispetto alle misurazioni dirette. L’analisi statistica dei dati raccolti ha permesso di confrontare i valori stimati con quelli ottenuti direttamente tramite biometro, valutando il grado di concordanza e identificando eventuali limiti di ciascun modello. Per ogni sottocategoria considerata, è stato determinato l’R2 e il bias medio tra valori calcolati e misurati. Si sono anche determinati i limiti di accordo al 95% tramite un’analisi di Bland-Altman per valutare la concordanza tra i due metodi di ottenimento dei dati di AL. la significatività dei risultati si è verificata mediante il calcolo dei p-value. In riferimento al campione totale, la formula proposta da Kim et al. ha prodotto bias compresi tra +1,36 mm e +1,65 mm (p < 0,01), con limiti di accordo ampi e R2 compresi tra 0,7905 e 0,9668 (p < 0,01), a seconda del tipo di SE considerato. Il modello di Morgan et al., invece, ha prodotto bias tra -0,11 mm e +0,05 mm (p > 0,05 per l’SE oggettivo), con limiti di accordo più ristretti rispetto al modello precedente e R2 tra 0,8263 e 0,8380 (p < 0,01). Infine, la formula di Queirós et al., ha fatto emergere bias in un intervallo di +0,05 e +0,18 (p > 0,05 per l’SE soggettivo), con limiti di accordo ristretti e R2 compresi tra 0,8284 e 0,8436 (p < 0,01). Inoltre, sono state eseguite ulteriori analisi per sottogruppi differenti di soggetti, che hanno confermato l’eterogeneità dell’applicabilità dei modelli. In seguito all’analisi, si è giunti alla conclusione per cui la predizione della lunghezza assiale tramite modelli matematici non rappresenta uno strumento sostitutivo alla biometria oculare propria, bensì complementare e che può essere applicato esclusivamente in contesti clinici non altamente rigorosi.
Verifica dell’affidabilità di formule predittive per la stima della lunghezza assiale, sulla base di parametri biometrici e refrattivi, in una popolazione di soggetti adulti.
BERTON, ELEONORA
2024/2025
Abstract
The purpose of the proposed cross-sectional study is to evaluate the effectiveness of some predictive models for axial length estimation, based on biometric data, through comparison with direct measurement made with the biometer itself. The study was conducted on a heterogeneous population of emmetropic and ametropic adult subjects. Exclusion criteria were established, including: positive history of trauma, surgery or current or past ocular disease that may affect biometry; children; subjects with unstable refraction; contact lens or orthokeratology lens wearers. The subjective refraction of the test population was determined by serving the phoropter while the objective refraction was obtained using the “i.Profiler” topographer and aberrometer. Biometric data collection was performed with the “Huvitz HBM-1” biometer, which provided measurements of axial length (AL) and corneal curvature (K1 and K2 with respective orientations). The data were transcribed into an Excel spreadsheet for processing. Consistent with the other studies analyzed, for each included sample, only the spherical equivalent was considered as the refractive reference. Three mathematical models were tested for the possibility of predicting axial length from corneal curvature and spherical equivalent, with subsequent comparison against direct measurements. Statistical analysis of the collected data allowed comparison of the estimated values with those obtained directly by biometer, assessing the degree of agreement and identifying any limitations of each model. For each subcategory considered, the R^2 and the average bias between calculated and measured values were determined. Limits of 95% agreement were also determined by a Bland-Altman analysis to assess the agreement between the two methods of obtaining AL data. the significance of the results was checked by calculating p-values. With reference to the total sample, the formula proposed by Kim et al. yielded biases ranging from +1.36 mm to +1.65 mm (p < 0.01), with wide limits of agreement and R2 ranging from 0.7905 to 0.9668 (p < 0.01), depending on the type of SE considered. In contrast, the Morgan et al. model produced bias between -0.11 mm and +0.05 mm (p > 0.05 for objective SE), with narrower limits of agreement than the previous model and R2 between 0.8263 and 0.8380 (p < 0.01). Finally, the Queirós et al. formula revealed bias in the range of +0.05 and +0.18 (p > 0.05 for subjective SE), with narrow limits of agreement and R2 between 0.8284 and 0.8436 (p < 0.01). In addition, further analyses were performed for different subgroups of subjects, which confirmed the heterogeneity of model applicability. As a result of the analysis, it was concluded that axial length prediction using mathematical models is not a substitute tool for proper ocular biometry, but rather a complementary one that can be applied only in clinical settings that are not highly rigorous.| File | Dimensione | Formato | |
|---|---|---|---|
|
Berton_Eleonora.pdf
Accesso riservato
Dimensione
1.42 MB
Formato
Adobe PDF
|
1.42 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/87619