Secondary Logo

Journal Logo


Popperian falsification of methods of assessing surgically induced astigmatism

Naeser, Kristian MDa,1,*

Author Information
Journal of Cataract & Refractive Surgery: January 2001 - Volume 27 - Issue 1 - p 25-30
doi: 10.1016/S0886-3350(00)00605-2
  • Free


In ophthalmologic literature, several conflicting methods for reporting surgically induced astigmatism (SIA) have been described.1–8 Some are incorrect,8,9 but as the observer has no real sense of what result to expect, the fallacy of a specific method may be difficult to conceive. A method of assessing SIA should be optically meaningful, consistent within its own definition, and correct for both single and aggregate data.

In the present study, I reviewed the various assessment methods and then tested the methods in situations in which the expected result was obvious. When this result was not produced, I regarded the specific assessment method as falsified in a classical “Popperian” manner. Karl Popper is known for his thesis that scientific theories cannot be proved but only falsified. Rather than attempting to elaborate on a theory, the scientist should constantly set up experiments allowing for falsification of the theory, thereby paving the way for new theories.

Materials and methods

Analysis of Aggregate Data

A calibrated steel test ball (Noco Nordisk Optisk Co.) of an assigned diameter of 15 mm was placed in its holder and mounted on the chin rest of a Haag-Streit slitlamp. Keratometry of the steel ball was performed twice with a Nidek model ARK-2000S autorefractokeratometer, as previously described.10 Following a double determination, the steel ball was rotated at random and a new double measurement performed. A total of 50 double measurements were performed.

In the following, the subscripts 1 and 2 refer to measurements derived from the first and second keratometer reading. The following entities were registered:

  • α1 and α2: Direction in degrees of the steeper corneal principal plane5
  • φ1 and φ2: Direction in degrees of the flatter corneal principal plane, where φ = α + 90°
  • M1 and M2: Difference in power, expressed in diopters (D), between the steeper and the flatter principal corneal plane5
  • (M1 @ α1) and (M2 @ α2): Net astigmatisms as previously described5,6

The mean difference between the measurements was calculated by several assessment methods, and 95% confidence intervals (95% CI) were computed for univariate data. Bivariate data were analyzed with the Hotelling T2 test.8,11 In all cases, the difference between the 2 measurements was taken as

For these data, the average difference between the 2 measurements would be expected to be close to and not statistically significantly different from zero, as the examined objects would not change as a result of the measurements. Assessment methods producing average differences significantly different from zero would be regarded as falsified.

Methods of assessing changes in net astigmatism have been described.12–19

Simple Subtraction Method

This method considers only the astigmatic magnitude, disregarding the astigmatic direction.

Algebraic Method4,15,20

M1alg and M2alg change signs in accordance with the direction of the steeper corneal meridian; for 45 ≤ α <135, a “+” sign is added; for all other values of α, M is denoted a minus sign. The method describes the with-the-rule (WTR) and against-the-rule (ATR) change between 2 measurements.

Cravy's Method2

where Δx = x1 – x2; in this equation x1 = M1 · cos α1, x2=M2 · cos α2, and Δy = y1 – y2; in this equation y1 = M1 · sin α1, y2 = M2 · sin α2. Schematically,

As in the algebraic method, the Cravy system reports the WTR and ATR change and has an elaborate convention2,4,15,18 for determining the delta values. Cravy's ΔK yields results very close to ΔKP(90).

Astigmatic Magnitude Not Considering Axis

Several methods have been described.1,3,4,15,16,21,22 All yield identical results and have various sign conventions.

Magnitude of induced astigmatism for a single patient or observation:

Direction of induced astigmatism for a single patient or observation:

For aggregate data, Mvec is averaged directly without considering direction.8,9,23

It is now generally agreed that separate analysis of astigmatic direction is meaningless.3,8,14 Average values for φvec will therefore not be reported.

Vector Decomposition

This method reports the WTR and ATR components from the values generated from astigmatic magnitude not considering axis.3,4,24

Polar Values5–10

Univariate analysis5–7,9:

The WTR and ATR components derived from polar values:

Bivariate statistical analysis was performed as previously described.8,10

Mean difference, expressed as net astigmatism, was calculated as previously described.7–9,24

Alpins25,26 and Holladay27 Methods

Both methods use the basal entities M · cos2 φ and M · sin2 φ and are therefore identical to the Bennett24 and polar value methods, shown in equations 8 and 9, as7,8

We will therefore not perform a separate analysis of these methods.

Analysis of Data from 2 Measurements from a Single Patient

I constructed examples in which the expected results were obvious. When this result was not produced by a specific assessment method, it was regarded as falsified.

Statistical Analysis

All data were processed and statistical analyses were performed with the SPSS program (Statistical Program for the Social Sciences). A 2-tailed paired t test was used for comparing double measurements on univariate data. Bivariate statistical analysis was performed with the Hotelling T2.8,11


Analysis of Aggregate Data

The various assessment methods of net astigmatisms yielded the mean differences between the 2 measurements shown in Table 1. The mean difference in astigmatism (0.02 D @ 173) was not significantly different from zero (bivariate analysis with the Hotelling T2 = 0.007, F = 0.164 with (2,48) degrees of freedom; P = .85). The simple subtraction, algebraic, Cravy, and polar value methods yielded results close to and not significantly different from zero. Conversely, astigmatic magnitude not considering axis and astigmatic decomposition both resulted in significant mean differences from zero. In this test situation for aggregate data, the vector analysis and astigmatic decomposition methods were falsified.

Table 1
Table 1:
Mean differences in astigmatism between 50 double keratometries on test steel balls, assessed by various methods.

Analysis of Data from a Single Patient

Example 1

Calculate the difference between the following 2 autokeratometry readings

As the magnitudes of the 2 net astigmatisms are identical and the directions are very close, little difference between the 2 measurements is expected. Table 2 shows the calculated values for the different methods.

Table 2
Table 2:
From example 1: The difference in astigmatism between 2 keratometries, assessed by various methods.

The simple subtraction, Cravy, vector analysis, astigmatic decomposition, and polar value methods yielded results close to zero. The algebraic method resulted in a large difference between the 2 measurements, falsifying this approach. This is naturally caused by the change in sign at 45 degrees.

Example 2

As the directions of the 2 meridians are orthogonal, the difference between measurement 1 and 2 easily calculates as (2.0 D @ 0). Results of the calculations are shown in Table 3.

Table 3
Table 3:
From example 2: The difference in astigmatism between 2 keratometries, assessed by various methods.

This example falsifies the simple subtraction method. The difference between the 2 astigmatisms, expressed as polar values in 135 degrees, also amounts to zero, but this is as expected as both net astigmatisms amount to zero for this specific polar value in the oblique meridian.

The joint analysis of aggregate and single data has falsified the simple subtraction, algebraic, astigmatic magnitude not considering axis, and astigmatic decomposition methods.


Previous studies of SIA have emphasized the mathematics involved. A different strategy was chosen for this study. A variety of assessment methods were tested under conditions in which the correct result was known or obvious. Assessment methods failing to produce this result were regarded falsified.

Astigmatism is characterized by both direction and magnitude, and it is therefore not surprising that the simple subtraction and algebraic methods, which consider only astigmatic magnitudes, are inconsistent, as shown in examples 1 and 2. As pointed out by Olsen and Dam-Johansen,14 the Cravy method is a mathematical concept, not based on optics. This method will not be considered in the following discussion.

Jaffe and Clayman1 originally described 2 methods for analysis of SIA. In one, the astigmatic direction was considered; in the other, it was disregarded.9 This last method has been extensively used under the term “Jaffe vector analysis.” Surgically induced astigmatism is calculated correctly for individual patients, while astigmatic direction is disregarded in the analysis of aggregate data.7–9,23,26,27 For average values, a systematic error is introduced, as astigmatisms in various degrees of inclination are not cancelled out.23 This group of methods1,3,4,15,16,21,22 was recently given the more neutral term astigmatic magnitude not considering axis.9 In the present study, the difference between 2 consecutive keratometry readings on unchanging surfaces averaged 0.19, a value significantly different from zero. Vector decomposition is based on the results from vector analysis and therefore also produced a systematic error (Table 1).

The polar value methods are conceptually based on the surgically induced flattening and torque of the preoperative cylinder. Alpins25,26 and Holladay27 have used different concepts to describe similar methods. The Naeser, Alpins, and Holladay methods agree with current research in optometry.28–30 The next steps are to take advantage of these valid methods fully by using them with multivariate statistics to describe not only astigmatism and spherocylinders8 but also more complex optical phenomena.


1. Jaffe NS, Clayman HM. The pathophysiology of corneal astigmatism after cataract extraction. Trans Am Acad Ophthalmol Otolaryngol 1975; 79:OP615-OP630
2. Cravy TV. Calculation of the change in corneal astigmatism following cataract extraction. Ophthalmic Surg 1979; 10:38-49
3. Holladay JT, Cravy TV, Koch DD. Calculating the surgically induced refractive change following ocular surgery. J Cataract Refract Surg 1992; 18:429-443
4. Dam-Johansen M, Olsen T, Theodorsen F. The long-term course of the surgically-induced astigmatism after a scleral tunnel incision. Eur J Implant Refract Surg 1994; 6:337-343
5. Naeser K. Conversion of keratometer readings to polar values. J Cataract Refract Surg 1990; 16:741-745
6. Naeser K, Behrens JK, Naeser EV. Quantitative assessment of corneal astigmatic surgery; expanding the polar values concept. J Cataract Refract Surg 1994; 20:162-168
7. Naeser K, Behrens JK. Correlation between polar values and vector analysis. J Cataract Refract Surg 1997; 23:76-81
8. Naeser K, Hjortdal JØ. Bivariate analysis of surgically induced regular astigmatism. Mathematical analysis and graphical display. Ophthalmic Physiol Opt 1999; 19:50-61
9. Naeser K. Assessment of surgically induced astigmatism; call for an international standard (letter). J Cataract Refract Surg 1997; 23:1278-1280
10. Naeser K, Guo S. Precision of autokeratometry expressed as confidence ellipses in Euclidian 2-space. Ophthalmic Physiol Opt 2000; 20:160-168
11. Hotelling H. The generalisation of Student's ratio. Ann Math Stat 1931; 2:360-378
12. Seiler T, Wollensak J. Über die mathematische Darstellung des postoperativen regulären Hornhautastigmatismus. Klin Monatsbl Augenheilkd 1993; 203:70-76
13. Toulemont PJ, Serdarevic ON. Modèles vectoriels d'analyses des variations d'astigmatism cornéen antérieur. J Fr Ophtalmol 1993; 16:472-481
14. Olsen T, Dam-Johansen M. Evaluating surgically induced astigmatism. J Cataract Refract Surg 1994; 20:517-522
15. Dam-Johansen M. Corneal astigmatisme efter cataract kirurgi (doctoral thesis). Århus, Denmark, Århus University, 1996
16. Jaffe NS, Jaffe MS, Jaffe GF. Cataract Surgery and Its Complications. 6th ed. St Louis, MO: Mosby, 1997; 132-146
17. Roman S, Dobigny M, Ullern M. Comment calculer un astigmatisme chirurgicalement induit. J Fr Ophtalmol 1997; 20:241-245
18. Nielsen PJ. Data collection and analysis. In: Yanoff M, Duker JS, eds., Ophthalmology. St. Louis, MO: Mosby, 1998; Chap 33, 1-6
19. Goes FM Jr, Missotten L. How to calculate surgically induced astigmatism after cataract surgery? Bull Soc Belge Ophtalmol 1998; 268:35-40
20. Richards SC, Brodstein RS, Richards WL, et al. Long-term course of surgically induced astigmatism. J Cataract Refract Surg 1988; 14:270-276
21. Naylor EJ. Astigmatic difference in refractive orders. Br J Ophthalmol 1968; 52:422-425
22. Kaye SB, Campbell SH, Davey K, Patterson A. A method for assessing the accuracy of surgical techniques in the correction of astigmatism. Br J Ophthalmol 1992; 76:738-740
23. Naeser K. Vector analysis may not apply for astigmatic surgery (letter). J Cataract Refract Surg 1993; 19:668-669
24. Bennett AG. A new approach to the statistical analysis of ocular astigmatism and astigmatic prescriptions. The Frontiers of Optometry: First International Congress 1984; 2:35-42
25. Alpins NA. A new method of analyzing vectors for changes in astigmatism. J Cataract Refract Surg 1993; 19:524-533
26. Alpins NA. Vector analysis of astigmatism changes by flattening, steepening, and torque. J Cataract Refract Surg 1997; 23:1503-1514
27. Holladay JT, Dudeja DR, Koch DD. Evaluating and reporting astigmatism for individual and aggregate data. J Cataract Refract Surg 1998; 24:57-65
28. Thibos LN, Wheeler W, Horner D. Power vectors: an application of Fourier analysis to the description and statistical analysis of refractive error. Optom Vis Sci 1997; 74:367-375
29. Harris WF. Dioptric power: its nature and its representation in three- and four-dimensional space. Optom Vis Sci 1997; 74:349-366
30. Harris WF. Astigmatism. Ophthalmic Physiol Opt 2000; 20:11-30
© 2001 by Lippincott Williams & Wilkins, Inc.