We read with interest the article by Shammas et al. on toric intraocular lens results.1 That various studies, including this one, have shown that it is preferable to rely on an assumption made by an algorithm, rather than on the actual measurement, is counter-intuitive, and reflects inaccuracy of the real assessment.2–5 However, from this, it cannot be concluded that a supposition about posterior astigmatism is better than its direct determination and devices with newer technologies will undoubtedly allow more reliable results to be obtained. The authors did not explicitly indicate what proportion of the eyes showed against-the-rule (ATR), with-the-rule (WTR), and oblique astigmatism in the posterior corneal surface. A percentage of eyes, as expected, had posterior WTR astigmatism, as shown in the supplemental figure published by the authors (which we modified including labels for posterior corneal steep meridian and highlighting those eyes with posterior ATR astigmatism) (Figure 1).Figure 1.: Corneas with posterior ATR astigmatism are included in the red shadow area.The centroid, which can be interpreted as a vectorial average of the astigmatisms, suffers from the same problems as an arithmetical average: it hides the magnitude of the scattered data, if they cancel each other out. That is why in the group analyzed by Shammas et al., both the preoperative and postoperative centroids were close to zero (0.10 diopter [D] and 0.08 D). It is therefore imperative not to analyze centroid in isolation, but also considering the average of the absolute value of the magnitude of astigmatism (which postoperatively was 0.42 ± 0.23 D for the entire series). When analyzing the prediction error using the Barrett toric algorithm, the centroid was 0.06 D, but the mean absolute magnitude was 0.42 ± 0.23 D, and the 95% confidence ellipse was clearly above 0.75 D in one of the meridians, with several eyes above 0.75 D of prediction error, which is undoubtedly clinically significant. It would be very useful to identify the source of the error in those eyes. An obvious cause is that the algorithm used assumes that all eyes have ATR posterior corneal astigmatism, which clearly is not the case. For this reason, it would be very interesting to divide the analysis of the performance of the algorithm into 2 groups: those with ATR posterior astigmatism and those in which this condition is not fulfilled. Finally, it is possible that devices based on optical coherence tomography dedicated to the anterior segment are more accurate in measuring posterior corneal astigmatism than those using a visible light slit with the Scheimpflug principle. Since the Argos biometer is based on a swept-source optical coherence tomography, we wonder why it does not use that technology to also measure posterior corneal curvature (as the IOLMaster700 based on similar technology does)? There is no doubt that we are still halfway to accurately correcting astigmatism with toric lenses, but it sounds completely logical that the way is to improve technologies to obtain a real consistent measurement of each cornea, so as not to have to rely on population assumptions, which may not apply to a particular case.