This study examined the repeatability of corneal tomographic parameters in patients with keratoconus with advanced corneal thinning in a real-world setting.
The results of the study demonstrate that the repeatability limits of all corneal tomography parameters increased with increased disease severity. Repeatability limits were higher in very severe keratoconus (TCT ≤ 330 µm) versus severe keratoconus (TCT 331–400 µm), which were collectively (sub-400 group; TCT ≤ 400 µm) higher than in eyes with moderate thinning (450-plus group; TCT 450–500 µm).(11) The absolute repeatability limits for Kmax and TCT in our study (Table 2) were higher than those reported by other investigators,(9,11,12,15–22,26) which ranged between 0.86–1.66 D and 8.26–28.15 µm, respectively (Supplementary Table 2). However, the repeatability limits of K1A, K2A, K1P, K2P and ASTIG in our study were similar to those reported in the literature, ranging between 0.51–2.08, 0.48–1.56, 0.27–0.55, 0.16–0.51 (Supplementary Table 3) and 1.05–1.93,(11,17) respectively. The severity of keratoconus was greater in our study and is likely to be one of the reasons for the increased repeatability limits of Kmax and TCT in our study.
Another reason for decreased repeatability in our study is the inclusion of scans without all “OK” QS values. When only scans with “OK” QS values were analysed, repeatability limits were reduced for all parameters compared to those with mixed QS value scores (Supplementary Table 1). This difference may be confounded by the lower severity of keratoconus in the QS value “OK” only group (mean Kmax and TCT of 63.82 D and 401.33 µm in this subgroup compared to 72.66 D and 355.63 µm in advanced keratoconus group that included all scans (sub-400 group)). Obtaining corneal tomography scans with a QS value of “OK” is often challenging in patients with advanced keratoconus due to fixation losses from poor visual acuity or intolerance to scans due to light sensitivity. Pentacam algorithms in more advanced keratoconus are also likely to perform more poorly – for both image acquisition and algorithmic parameter generation. When acquiring images, the auto-align feature of the device often cannot detect where to scan, necessitating the need to align and initiate scans manually. Furthermore, corneas tend to be more scarred in advanced disease, which can impair the consistent reading of tissue boundaries and thus affect how algorithms read measurements and generate numeric values.
Nevertheless, scans can be of adequate quality if there are enough data points to generate a tomography map. Thus this study did not exclude scans without a QS value of “OK” for its primary analyses. While obtaining scans with a QS value of “OK” would be optimal, it is not always practical. Therefore, we propose that scans should be assessed qualitatively and interpreted in the context of the degree of variability. Alternatively, using mean measurements from multiple scans(27) may be useful in advanced keratoconus instead of relying solely on the QS value and a single scan.
The repeatability limits of different tomographic parameters quantified in this study have important implications for both the application of CXL in patients with advanced keratoconus as well as monitoring progression of disease. Parameters such as TCT are pivotal when deciding whether to offer CXL to patients with keratoconus. Knowing the repeatability limits reported in our study can help practitioners make informed decisions about the type of CXL protocol utilized from the beginning of the CXL procedure. For example, in patients with corneas bordering a TCT of 400 µm, practitioners may choose to utilise hypoosmolar riboflavin(28) or a sub-400 protocol(6) from the outset than switching mid-procedure.
Our study found that K1A, K2A, K1P and K2P had equivalent or better repeatability than TCT in both the sub-400 and 450-plus groups. This can be explained by the fact that these parameters represent more global measurements of the corneal surface rather than single-point parameters like TCT and thus, are expected to be more repeatable. Therefore, it is more logical to utilise newer progression analysis systems, such as ABCD,(29) that incorporate more repeatable parameters.
This study had several limitations, including differences in contact lens wear between the sub-400 and 450-plus groups, as well as the lack of inter-investigator reproducibility assessment. Firstly, corneal warpage can take one to eight weeks to stabilise.(30–32) However, it is often impractical to request patients with advanced keratoconus, who are often reliant on contact lenses, to be free of contact lenses for such prolonged periods. Secondly, the effects of any residual corneal warpage is likely to be consistent across scans, as scans were conducted within a 15 minute timeframe. Thus, while absolute keratometry and pachymetry values may be affected, repeatability measures are unlikely to be significantly affected. Secondly, while the repeatability of TCT is likely more important than reproducibility in surgical planning, the assessment of progression is likely to be influenced more by reproducibility as it is assessed over time and potentially by different examiners.(11) Future studies should investigate the reproducibility of corneal tomography parameters using the Pentacam in keratoconus.
In conclusion, this study quantified the repeatability limits of corneal tomographic parameters in keratoconic patients with moderate and advanced corneal thinning, which has implications for the safe application of CXL and identifying disease progression. When assessing patients with advanced thinning (TCT ≤ 400 µm), clinicians should strive to obtain scans with “OK” quality value if possible and, if unable after several attempts, assess the variability between scans to discern the most likely values of parameters. Utilising the mean of several measurements – ideally after acquiring a minimum of three scans – may also be helpful. Practitioners should utilise newer progression analysis systems that incorporate more repeatable parameters.