Figure 2a shows lens powers calculated using the Bennett method with customized
c constants calculated for the combination of the two datasets (184 eyes) as a function of axial length
L. The lens power has a negative correlation with axial length for
L < 24 mm (
r = –0.624;
P < 0.001), with a slope that matches that of the measured lens power data. Above
L = 24 mm, approximately corresponding with the onset of myopia, the lens power plateaus to become constant (
r = –0.036;
P > 0.05). Because phakometry was not available for the second dataset, this plateauing could not be confirmed experimentally. However, a similar trend was found in the raw data published by Sorsby et al.
17 Thus, in the absence of phakometry data for the entire dataset, the Bennett power with customized
c constants was used as a benchmark. This choice is based on the observation of Dunne et al.
8 that the Bennett power corresponds well with phakometry in myopic refractions up to –9.37 D, including the long eyes for which the plateauing is shown in
Figure 2a.
The mean powers with the modified-Stenström and Bennett-Rabbetts methods, using the Gullstrand-Emsley and Bennett-Rabbetts eye models, were 0.5 to 1.0 D less than the mean powers obtained with the Bennett method and its customized
c constants (
Table 2). These differences were statistically significant (paired
t-test,
P < 0.001). Using the customized
c constants, the modified-Stenström and Bennett-Rabbetts methods each yielded lens powers that were 0.71 ± 0.56 D greater than those with the Bennett method (
Table 3), and this difference was also statistically significant (
P < 0.001).
To improve the matches of the modified-Stenström and Bennett-Rabbetts methods with the Bennett method, a second
c constant (customized 2) was determined for the modified-Stenström and Bennett-Rabbetts methods that minimized the mean lens power difference with the Bennett method over the entire population. When these customized 2 constants were used, the lens power differences with the Bennett method were no longer statistically significant (
P > 0.05) and were within ±1 D for about 95% of eyes (
Table 3). For both methods, the power differences with the Bennett method were correlated significantly with axial length
L (
r = 0.390,
P < 0.001 and
r = 0.329,
P < 0.001 for the modified-Stenström and the Bennett-Rabbetts methods, respectively;
Fig. 2b).