Skip to main content
. 2023 Nov 15;2023(11):CD014911. doi: 10.1002/14651858.CD014911.pub2

Kamiya 2021.

Study characteristics
Patient Sampling Retrospective, single‐centre (Japan), case‐control study involving 349 keratoconus eyes and 170 normal eyes (refractive surgery candidates, contact lens fitting candidates).
Patient characteristics and setting A total of 349 eyes with good‐quality images of corneal topography measured with a Placido disc corneal topographer (TMS‐4 TM, Tomey, Aichi, Japan) were included. The disease was graded according to the Amsler‐Krumeich classification, as follows.
  • Grade 1 (54 eyes)

  • Grade 2 (52 eyes)

  • Grade 3 (23 eyes)

  • Grade 4 (50 eyes)


Control group: 170 eyes in people with normal ocular findings applying for a contact lens fitting or for a refractive surgery consultation, who had a refractive error of < 6 D as well as astigmatism of < 3 D.
Index tests Deep learning (convolutional neural network) of a single colour‐coded topography map.
Target condition and reference standard(s) Multiple corneal specialists diagnosed keratoconus with distinctive features (e.g. corneal colour‐coded map with asymmetric bowtie pattern with or without skewed axes), and ≥ 1 keratoconus sign (e.g. stromal thinning, conical bulging, Fleischer's ring, Vogt's striae, or apical scar). It is unclear how many corneal specialists classified the cases; however, classification was performed before inclusion.
Flow and timing All cases were included in the reference standard and index test. All data were included in a 2 × 2 table.
Comparative Not applicable
Notes This work was in part supported by Grants‐in‐Aid for Scientific Research (Grant Number 21K09706).
Methodological quality
Item Authors' judgement Risk of bias Applicability concerns
DOMAIN 1: Patient selection
Was a consecutive or random sample of patients enrolled? No    
Was a case‐control design avoided? No    
Did the study avoid inappropriate exclusions? Unclear    
Could the selection of patients have introduced bias?   High risk  
Are there concerns that the included patients and setting do not match the review question?     High
DOMAIN 2: Index test (All tests)
Were the index test results interpreted without knowledge of the results of the reference standard? Yes    
If a threshold was used, was it pre‐specified? Yes    
Was the model designed in an appropriate manner? Unclear    
Could the conduct or interpretation of the index test have introduced bias?   Low risk  
Are there concerns that the index test, its conduct, or interpretation differ from the review question?     Low concern
DOMAIN 3: Reference standard
Is the reference standard likely to correctly classify the target condition? Yes    
Were the reference standard results interpreted without knowledge of the results of the index tests? Yes    
Could the reference standard, its conduct, or its interpretation have introduced bias?   Low risk  
Are there concerns that the target condition as defined by the reference standard does not match the question?     Low concern
DOMAIN 4: Flow and timing
Did all patients receive the same reference standard? Yes    
Were all patients included in the analysis? Yes    
Could the patient flow have introduced bias?   Low risk  
DOMAIN 5: Comparative