Jump to comment:

• Calculation errors overemphasise the value of increasing visual field test frequency.
  Andrew Carkeet
  Published on: 6 April 2023
• Authors' Reply
  Balwantray C Chauhan
  Published on: 4 September 2008
• Is a field every 4 month a significant improvement over a field every 6 months?
  Albert Alm
  Published on: 28 July 2008

• Published on: 6 April 2023

  Calculation errors overemphasise the value of increasing visual field test frequency.

  • Andrew Carkeet, Senior Lecturer QUT School of Optometry and Vision Science

  Chauhan and co-workers [1] have provided Table 1, showing times taken to detect significant field progression with 80% power, based on a number of modelling parameters: frequency of examinations, rate of field progression, intrasession variability of field assessment. They have also provided Table 2 showing the number of annual eye examinations required to detect different total visual field changes, for different time periods, and for moderate variability. I have checked the calculations of Chauhan and co-workers, using Monte Carlo modelling, assuming a one-tailed significance value of 0.025. Of the 36 outcome values in Table 1, 33 are incorrect. Of the 12 outcome values in Table 2, 11 are incorrect.

  Chauhan and co-workers have made 2 main errors in their calculations for Table 1. The first is in applying their estimates of power. The curves shown in Figure 2 (statistical power plotted against number of field examinations) are appropriate for the case of 2 field examinations per year, but Chauhan and co-workers appear to have incorrectly also used them for the cases of 1 examination per year and 3 examinations per year. Separate sets of curves should have been calculated for those conditions. The effect on Table 1 is that the time taken to detect a field change is incorrectly reported as being inversely proportional to the number of examinations per year. This anomalous relationship was commented on by Albert Alm in his 2008 Rapid Response, “Is a field every 4...

  Show More

  Chauhan and co-workers [1] have provided Table 1, showing times taken to detect significant field progression with 80% power, based on a number of modelling parameters: frequency of examinations, rate of field progression, intrasession variability of field assessment. They have also provided Table 2 showing the number of annual eye examinations required to detect different total visual field changes, for different time periods, and for moderate variability. I have checked the calculations of Chauhan and co-workers, using Monte Carlo modelling, assuming a one-tailed significance value of 0.025. Of the 36 outcome values in Table 1, 33 are incorrect. Of the 12 outcome values in Table 2, 11 are incorrect.

  Chauhan and co-workers have made 2 main errors in their calculations for Table 1. The first is in applying their estimates of power. The curves shown in Figure 2 (statistical power plotted against number of field examinations) are appropriate for the case of 2 field examinations per year, but Chauhan and co-workers appear to have incorrectly also used them for the cases of 1 examination per year and 3 examinations per year. Separate sets of curves should have been calculated for those conditions. The effect on Table 1 is that the time taken to detect a field change is incorrectly reported as being inversely proportional to the number of examinations per year. This anomalous relationship was commented on by Albert Alm in his 2008 Rapid Response, “Is a field every 4 month a significant improvement over a field every 6 months?”. He wrote, “… increasing the duration is much more efficient than increasing the frequency of examinations. Thus, 5 examinations in 1.7 years will not be able to detect the same slope as 5 examinations in 5 years!” My calculations for the values in Table 1 show that increasing examination frequency, from once per year to 3 times per year, reduces time to detect a field change to between 71% and 50%, not 33% as implied by the original paper.

  The second error made by Chauhan and co-workers is in converting from the number of fields (S) required to the duration required (D) for a given annual frequency of testing (f). From equation 5 in another paper,[2] this relationship is: D=(S-1)/f. Chauhan and co-workers appear to have used D=S/f.
  In Table 1, three values are correct. One of those correct values, that it takes 2 years to detect fast progression (MD rate =-2dB/Year) with moderate field variability (SD=1dB) with 3 examinations/year, seems to be the basis for the paper’s recommendation that “six visual field examinations should be performed in the first 2 years.” This is slightly incorrect. A testing frequency of 3 examinations per year over 2 years will yield 7 examinations.

  Table 2 contains significant errors. For example, its results imply that detecting a total field change of 1 dB will require between 15 and 21 measurements with SD= 1dB, depending on the number of years measurements are spread over. My calculations give an estimate of 95 total measurements required (irrespective of the time span).
  This paper has been, and continues to be, influential. For example, Table 1 has been included in clinical glaucoma management guidelines.[3] The journal’s article metrics show 371 citations for the paper at the time of writing. Yet, the paper incorrectly overemphasises the value of increasing visual field test frequency. That is a core message for the paper. Given the paper’s influence those errors should be corrected.

  References
  1. Chauhan BC, Garway-Heath DF, Goñi FJ, et al. Practical recommendations for measuring rates of visual field change in glaucoma. Br J Ophthalmol 2008;92(4):569-73. doi: 10.1136/bjo.2007.135012 [published Online First: 2008/01/24]
  2. Schlesselman JJ. Planning a longitudinal study. II. Frequency of measurement and study duration. J Chronic Dis 1973;26(9):561-70. doi: 10.1016/0021-9681(73)90061-1
  3. NHMRC. NHMRC Guidelines for the screening, prognosis, diagnosis, management and prevention of glaucoma 2010: National Health and Medical Research Council, Canberra., 2010.

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top
• Published on: 4 September 2008

  Authors' Reply

  • Balwantray C Chauhan
  • Other Contributors:
    • David F. Garway-Heath, Francisco J. Goñi, Luca Rossetti, Boel Bengtsson, Ananth C Viswanathan, and Anders Heijl

  Dear Editor

  We thank Dr. Alm for his interest and comment on our paper entitled “Practical recommendations for measuring rates of visual field change in glaucoma.” We agree that the standard error of slope estimates is dependent on the number of examinations and duration of follow-up. However, these two parameters are not interchangeable. As pointed out correctly by Dr. Alm, the same number of examinations ov...

  Show More

  Dear Editor

  We thank Dr. Alm for his interest and comment on our paper entitled “Practical recommendations for measuring rates of visual field change in glaucoma.” We agree that the standard error of slope estimates is dependent on the number of examinations and duration of follow-up. However, these two parameters are not interchangeable. As pointed out correctly by Dr. Alm, the same number of examinations over a longer time period will lead to a better slope estimation and therefore greater power simply because the amount of net change is larger. Hence the difference in follow-up of 3 (27 compared to 24 months) months in his example yields an additional change of -0.5 dB in Mean Deviation and therefore the number of examinations derived to obtain equivalent power is not surprising.

  The goal of our first recommendation of 6 examinations in the first two years was not to determine whether the slope of visual field decay is statistically significant or not. It was to identify patients who we feel are progressing rapidly. In these patients we do not feel that prolonging the follow-up is an adequate substitute for more frequent examinations. Even if equivalent statistical power is obtained with this approach, prolonging the follow-up comes with the very significant cost of more visual field loss.

  Our paper was meant to be an initial guideline for clinicians on what rates of visual field change could be detected (with varying degrees of power) with a given frequency of examinations. We would like to emphasise that there is a large distinction between detection of progression and determining the rate of visual field change. Therefore the proposed guidelines are not necessarily designed to detect progression. We welcome the comments of Dr. Alm and others that encourage us to factor in aspects such as the time-separation between tests and measures besides Mean Deviation which consider the location of visual field defects.

  Sincerely,

  BC Chauhan, DF Garway-Heath, FJ Goñi, L Rossetti, B Bengtsson, AC Viswanathan and A Heijl

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top
• Published on: 28 July 2008

  Is a field every 4 month a significant improvement over a field every 6 months?

  • Albert Alm, Professor emeritus

  Dear Editors

  How often should we do visual fields in the first 2 years? Chauhan and co-workers [1] recommend 3 visual fields per year. It will have an 80% power of detecting a rate of loss of 2 dB/year in an eye with moderate variability. Is this a significant improvement over 2 fields per year? In order to answer that we should look at the efficacy of increasing the frequency of field examinations versus prolongin...

  Show More

  Dear Editors

  How often should we do visual fields in the first 2 years? Chauhan and co-workers [1] recommend 3 visual fields per year. It will have an 80% power of detecting a rate of loss of 2 dB/year in an eye with moderate variability. Is this a significant improvement over 2 fields per year? In order to answer that we should look at the efficacy of increasing the frequency of field examinations versus prolonging the duration of the follow-up. From regression theory it is clear that the ability to detect a statistically significant slope is determined by the variability of measurement (MD), the number of observations/examinations and the span along the X-axis (duration of follow-up). Furthermore, increasing the duration is much more efficient than increasing the frequency of examinations. Thus, 5 examinations in 1.7 years will not be able to detect the same slope as 5 examinations in 5 years! The total amount of change (rate * duration) will be much lower, and, as pointed out by the authors, the amount of change affects power. Planning of a longitudinal study is a useful parallel. Extensive tables, based on regression theory, on the effect of number of examinations and duration of the study on the precision of the estimate of the linear regression line have been published.[2]From these tables it can be seen that five measurements, evenly spread over 27 months, have the same power to detect a statistically significant slope as seven measurements over 24 months. Thus a visual field every 6 month’ the first two years would seem to be an optimal frequency, if the goal is to determine if the slope of the regression line is statistically significant or not.

  References
  [1] Chauhan BC, Garway-Heath DF, Goñi FJ, Rossetti L, Bengtsson B, Viswanathan AC, and Heijl A. Practical recommendations for measuring rates of visual field change in glaucoma. Br J Ophthalmol. 2008;92:569-73.
  [2] Schlesselman JJ. Planning a longitudinal study: II. Frequency of measurement and study duration. J Chron Dis. 1973;26:561-570.

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top