Validity and reliability of ankle morphological measurements on computerized tomography-synthesized planar radiographs

Background: Clinical success of total ankle arthroplasty depends heavily on the available information on the morphology of the bones, often obtained from measurements on planar radiographs. The current study aimed to evaluate the intra-rater, inter-rater and inter-session reliability and the validity of radiograph-based measurements of ankle morphology, and to quantify the effects of examiner experience on these measurements. Methods: Twenty-four fresh frozen ankle specimens were CT scanned, data of which were used to reconstruct 3D volumetric bone models for synthesizing 2D radiographs. Two orthopaedic surgeons with different levels of clinical experience identified twenty landmarks five times on each of the synthesized sagittal and coronal radiographs and repeated the test on a subsequent day within 5 days. The landmarks were used to calculate fourteen morphological parameters. The two-way mixed-effects (ICC3,1), two-way random-effects (ICC2,k) and two-way random-effects (ICC3,k) models were used, respectively, to assess the intra-rater, inter-rater and inter-session reliability of measurements. The validity of the measurements for each examiner was assessed by comparing them with gold standard values obtained from the 2D radiographs projected from the 3D volumetric models using Pearson's correlation analysis and Bland and Altman plots, and the differences were defined as the measurement errors. Results: Most of the morphological parameters were of good to very good intra-rater, inter-session and inter-rater reliability for both examiners (ICC > 0.61). Experience appeared to affect the inter-rater and inter-session reliability, the senior examiner showing greater inter-session ICC values than the junior examiner. Most of the tibial parameters had moderate to excellent correlations with the corresponding gold standard values but were underestimated by both examiners, in contrast to most of the talar parameters that were overestimated and had only poor to fair correlations. Conclusions: Most of the morphological parameters of the ankle can be estimated from radiographs with good to very good intra-rater, inter-session and inter-rater reliability, for both clinically experienced and less experienced examiners. Clinical experience helped increase the reliability of repeated evaluations after a longer interval, such as in a follow-up assessment. It is suggested that critical clinical decisions based on repeated morphology measurements should be made by more experienced surgeons or after appropriate training