Endothelial cell density and morphometry are essential for adequate follow-up of corneal grafts and diseases. Fast, easy-to-use and reliable instruments to measure the endothelial cell layer are required in clinical routine. While contact instruments provide excellent images, these have the disadvantage of directly touching the cornea. Non-contact specular microscopes are appreciated by clinicians and patients for their convenient handling. In this study we focussed on two non-contact instruments from Konan and Topcon, two widely used non-contact specular microscope manufactors in this field.
The models used in this study are the Konan Noncon Robo SP8000 from the Konan Robo SP series and the Topcon SP3000P from the Topcon SP-series, which provide high magnification views of specular reflected light from the corneal endothelium. Both offer auto-alignment to capture the images. Several methods of cell analysis can be applied: Manual, semi-automatic or automatic cell counting strategies. While automatic strategies are appealing because of ease of use, these are known to be less accurate than semi-manual or manual cell detection [5]. Thus we used manual cell detection for both instruments in the present study.
In addition, the number of marked cells that are incorporated into the cell density calculation can influence the results for cell density [6–8]. It is advised to dot as many cells as possible, since the more cells are included in the analysis the smaller the resulting variations [9]. In the present study, we therefore marked and thus included into the calculations as many cells as possible for the two devices: All clearly visible cells on the picture for the Konan Robo instrument were selected, and for the Topcon device, cells were manually marked until the built-in software ceased further cell marking. These methods resulted in a tendency for a higher number of analyzed cells for the Topcon instrument without statistical significance.
As it is also known that inter-observer variation can occur, all examinations and cell dotting with both instruments were performed by the same examiner.
Other study limitations are the cohort size. As the performance of measuring instruments und thus the congruence of two machines might be dependent on cell density and regularity, we included not only healthy, untreated corneas but a wide range of patients to include also a wide range of endothelial cell mosaics into our comparison. The analysis covers endothelial cell counts from as low as about 600 cells/mm2 to around 3000 cells/mm2, with around one third of measurements below 2000 cells/mm2. According to a linear regression model to declare influencing factors on the measurement agreement for the two instruments, neither a low ECD nor age turned out statistically significant predictors of the difference between endothelial cell measurements of both instruments. Nonetheless, it is possible that in special situations like e.g. certain diseases as cornea guttata, or poor image quality (which was not included in our study) the observed difference between the instruments might be altered. However, even with a limited number of eyes examined, and a wide range of cell densities taken into account, our study shows that results are rather consistent: In 94 % of examinations, Konan gave higher cell counts for ECD than Topcon.
Since ECD measurements are mostly used for follow-up of endothelial cell changes in individual patients or for clinical studies, it is important to know if or which instruments can be used interchangeably. Several studies have compared endothelial cell counts from models of the Topcon SP series with other contact [10] or non-contact microscopes [11, 12]. Thuret et al. compared the Topcon SP2000P with a non-contact specular microscope from Rhine-Tec, Germany. When using the semi-automatic mode for both instruments, agreement was far better than compared to the automatic mode, but still the Rhine-Tec showed the tendency to overestimate low and underestimate high endothelial cell densities compared to the Topcon [12]. De Sanctis et al. compared the same Topcon instrument SP2000P with the Konan CC7000 non-contact specular microscope. In this analysis, the endothelial cell densities measured by the Konan where statistically significant higher than those calculated by the Topcon; depending on the examiner, the mean difference in ECD was 185 to 229 cells between the two instruments [11]. In our analysis, we used different models, but instruments from the same manufactorers, and found similar results: We found higher ECD for our Konan model compared to the Topcon instrument. However, we also included patients with corneal disease or following keratoplasty to analyze a wider range of ECDs, while in the study by de Sanctis et al. only untreated healthy subjects where enrolled. In addition, they used the semi-automatic mode and marked 75–88 or 80 cells, respectively, while we used the manual mode and marked a mean of 100 or 110 cells respectively per image to minimize discrepancies caused by calculation inaccuracies due to low numbers of included cells. Apart from the mentioned study, Konan non-contact microscopes have also been compared to contact microscopes for ECD measurements [13, 14]. When comparing the Konan Robo SP8000 to a non-contact instrument by Zeiss regarding mean cell area, significant differences where detected so that the authors recommend not to use the instruments interchangeably [13]. Several studies have shown that results for ECD, hexagonality or cell polymorphism can differ significantly. Luft et al. compared a Konan instrument (CellChek XL) to three non-contact models from other manufacturers (Bon Optics, Tomey and Nidek) both in healthy and compromised corneas. They found little consistency between the 4 devices with respect to the qualitative endothelial cell parameters CV and hexagonality readings [15].
Due to the discrepancies in all parameters tested in our own study, we recommend not to use the Konan and the Topcon interchangeably in the same patient.