1 of 51

Proposal of a Simpler Risk Model Incorporating RPD

for the Clinical Prediction of Late AMD

by Matt Trinh

PhD, M Optom, B Optom / B Sc (Hons)

m.trinh@unsw.edu.au

Co-authors: Annita Duong, Rene Cheung, Simon Chen, David Ng, Jeff Friedrich, Chris Hodge, Lisa Nivison-Smith, Angelica Ly

Posterior Eye Education & Research Society

2 of 51

Acknowledgements

Some graphics in this presentation generated using ChatGPT-4o.

No parts of the study used large language models.

Thank you to the Future Vision Foundation (Australia) for funding support – seed grant.
Thank you to the ARVO foundation for funding support – travel grant.

Posterior Eye Education & Research Society

3 of 51

Why is AMD prognostication important?

By forming an accurate prognosis and stratifying pxs by risk, we can:

Optimise clinical management, i.e., efficient monitoring and earlier intervention

Posterior Eye Education & Research Society

4 of 51

Why is AMD prognostication important?

By forming an accurate prognosis and stratifying pxs by risk, we can:

Optimise clinical management, i.e., efficient monitoring and earlier intervention

Enhance study design, i.e., targeted enrolment and risk-adjusted analyses

Posterior Eye Education & Research Society

5 of 51

Current AMD prognostication

Clinical standard for simplified AMD risk predictions^*
(Ferris et al. 2005, AREDS report #18)
Predicts progression to late AMD by person

2 key biomarkers (drusen size and pigmentary abnormalities)

*According to the American Academy of Ophthalmology Preferred Practice Pattern for AMD

Ferris, F.L. et al. (2005) ‘A simplified severity scale for age-related macular degeneration: AREDS report no. 18’, Archives of Ophthalmology, 123(11), pp. 1570–1574. Available at: https://doi.org/10.1001/archopht.123.11.1570.

Posterior Eye Education & Research Society

6 of 51

Current AMD prognostication

Clinical standard for simplified AMD risk predictions^*
(Ferris et al. 2005, AREDS report #18)
Predicts progression to late AMD by person

2 key biomarkers (drusen size and pigmentary abnormalities)

Up to 7 total observations across both eyes

1-2) Large drusen …in each eye

3-4) Pigmentary abnormalities …in each eye

5) Late AMD …in fellow eye

6-7) Intermediate drusen …in both eyes (if large drusen not present)

*According to the American Academy of Ophthalmology Preferred Practice Pattern for AMD

Ferris, F.L. et al. (2005) ‘A simplified severity scale for age-related macular degeneration: AREDS report no. 18’, Archives of Ophthalmology, 123(11), pp. 1570–1574. Available at: https://doi.org/10.1001/archopht.123.11.1570.

Posterior Eye Education & Research Society

7 of 51

Current AMD prognostication

Clinical standard for simplified AMD risk predictions^*
(Ferris et al. 2005, AREDS report #18)
Predicts progression to late AMD by person

2 key biomarkers (drusen size and pigmentary abnormalities)

Up to 7 total observations across both eyes

1-2) Large drusen …in each eye

3-4) Pigmentary abnormalities …in each eye

5) Late AMD …in fellow eye

6-7) Intermediate drusen …in both eyes (if large drusen not present)

5 annual risk scores (0-to-4)

E.g., 5yrs 🡪 ≈0.5%, 3%, 12%, 25%, 50%

*According to the American Academy of Ophthalmology Preferred Practice Pattern for AMD

Ferris, F.L. et al. (2005) ‘A simplified severity scale for age-related macular degeneration: AREDS report no. 18’, Archives of Ophthalmology, 123(11), pp. 1570–1574. Available at: https://doi.org/10.1001/archopht.123.11.1570.

Posterior Eye Education & Research Society

8 of 51

Current AMD prognostication

Clinical standard for updated simplified AMD risk predictions^*

(Agrón et al. 2024, AREDS report #42)

Predicts progression to late AMD by person

3 key biomarkers (drusen size, pigmentary abnormalities, RPD)

Up to 9 total observations across both eyes

1-2) Large drusen …in each eye

3-4) Pigmentary abnormalities …in each eye

5) Late AMD …in fellow eye

6-7) Intermediate drusen …in both eyes (if large drusen not present)

8-9) RPD …in either eye

10 annual risk scores (0-to-4, without and with RPD)

Score: 0 1 2 3 4

E.g. 5yrs without RPD 🡪 ≈0.5%, 3%, 12%, 25%, 50%
E.g., 5yrs with RPD 🡪 3%, 8%, 29%, 59%, 72%

*According to the American Academy of Ophthalmology Preferred Practice Pattern for AMD

Ferris, F.L. et al. (2005) ‘A simplified severity scale for age-related macular degeneration: AREDS report no. 18’, Archives of Ophthalmology, 123(11), pp. 1570–1574. Available at: https://doi.org/10.1001/archopht.123.11.1570.

Agrón, E. et al. (2024) ‘An updated simplified severity scale for age-related macular degeneration incorporating reticular pseudodrusen: Age-Related Eye Disease Study report number 42’, Ophthalmology, 131(10), pp. 1164–1174. Available at: https://doi.org/10.1016/j.ophtha.2024.04.011.

Posterior Eye Education & Research Society

9 of 51

Current AMD prognostication

Clinical standard for updated simplified AMD risk predictions^*

(Agrón et al. 2024, AREDS report #42)

Predicts progression to late AMD by person

3 key biomarkers (drusen size, pigmentary abnormalities, RPD)

Up to 9 total observations across both eyes

1-2) Large drusen …in each eye

3-4) Pigmentary abnormalities …in each eye

5) Late AMD …in fellow eye

6-7) Intermediate drusen …in both eyes (if large drusen not present)

8-9) RPD …in either eye

10 annual risk scores (0-to-4, without and with RPD)

Score: 0 1 2 3 4

E.g. 5yrs without RPD 🡪 ≈0.5%, 3%, 12%, 25%, 50%
E.g., 5yrs with RPD 🡪 3%, 8%, 29%, 59%, 72%

*According to the American Academy of Ophthalmology Preferred Practice Pattern for AMD

Ferris, F.L. et al. (2005) ‘A simplified severity scale for age-related macular degeneration: AREDS report no. 18’, Archives of Ophthalmology, 123(11), pp. 1570–1574. Available at: https://doi.org/10.1001/archopht.123.11.1570.

Agrón, E. et al. (2024) ‘An updated simplified severity scale for age-related macular degeneration incorporating reticular pseudodrusen: Age-Related Eye Disease Study report number 42’, Ophthalmology, 131(10), pp. 1164–1174. Available at: https://doi.org/10.1016/j.ophtha.2024.04.011.

Posterior Eye Education & Research Society

10 of 51

Potential tweaks

Clinical standard for updated simplified AMD risk predictions^*

(Agrón et al. 2024, AREDS report #42)

Predicts progression to late AMD by person

3 key biomarkers (drusen size, pigmentary abnormalities, RPD)

Up to 9 total observations across both eyes

1-2) Large drusen …in each eye

3-4) Pigmentary abnormalities …in each eye

5) Late AMD …in fellow eye

6-7) Intermediate drusen …in both eyes (if large drusen not present)

8-9) RPD …in either eye

10 annual risk scores (0-to-4, without and with RPD)

Score: 0 1 2 3 4

E.g. 5yrs without RPD 🡪 ≈0.5%, 3%, 12%, 25%, 50%
E.g., 5yrs with RPD 🡪 3%, 8%, 29%, 59%, 72%

*According to the American Academy of Ophthalmology Preferred Practice Pattern for AMD

Ferris, F.L. et al. (2005) ‘A simplified severity scale for age-related macular degeneration: AREDS report no. 18’, Archives of Ophthalmology, 123(11), pp. 1570–1574. Available at: https://doi.org/10.1001/archopht.123.11.1570.

Agrón, E. et al. (2024) ‘An updated simplified severity scale for age-related macular degeneration incorporating reticular pseudodrusen: Age-Related Eye Disease Study report number 42’, Ophthalmology, 131(10), pp. 1164–1174. Available at: https://doi.org/10.1016/j.ophtha.2024.04.011.

1) Reduce the number of observations?

Posterior Eye Education & Research Society

11 of 51

Potential tweaks

Clinical standard for updated simplified AMD risk predictions^*

(Agrón et al. 2024, AREDS report #42)

Predicts progression to late AMD by person

3 key biomarkers (drusen size, pigmentary abnormalities, RPD)

Up to 9 total observations across both eyes

1-2) Large drusen …in each eye

3-4) Pigmentary abnormalities …in each eye

5) Late AMD …in fellow eye

6-7) Intermediate drusen …in both eyes (if large drusen not present)

8-9) RPD …in either eye

10 annual risk scores (0-to-4, without and with RPD)

Score: 0 1 2 3 4

E.g. 5yrs without RPD 🡪 ≈0.5%, 3%, 12%, 25%, 50%
E.g., 5yrs with RPD 🡪 3%, 8%, 29%, 59%, 72%

*According to the American Academy of Ophthalmology Preferred Practice Pattern for AMD

Ferris, F.L. et al. (2005) ‘A simplified severity scale for age-related macular degeneration: AREDS report no. 18’, Archives of Ophthalmology, 123(11), pp. 1570–1574. Available at: https://doi.org/10.1001/archopht.123.11.1570.

Agrón, E. et al. (2024) ‘An updated simplified severity scale for age-related macular degeneration incorporating reticular pseudodrusen: Age-Related Eye Disease Study report number 42’, Ophthalmology, 131(10), pp. 1164–1174. Available at: https://doi.org/10.1016/j.ophtha.2024.04.011.

2) Reduce the number of risk scores?

1) Reduce the number of observations?

Posterior Eye Education & Research Society

12 of 51

Purpose and methods

Purpose

To develop an abridged risk model for AMD to:

Enable inter-eye risk comparisons
Potentially enhance clinical efficiency

Posterior Eye Education & Research Society

13 of 51

Purpose and methods

Purpose

To develop an abridged risk model for AMD to:

Enable inter-eye risk comparisons
Potentially enhance clinical efficiency

Methods (sampling)

Retrospective cohort study (n = 269) of pxs with early/intermediate AMD over seven-years
From 3x private ophthal practices in Sydney, Australia

Posterior Eye Education & Research Society

14 of 51

Purpose and methods

Purpose

To develop an abridged risk model for AMD to:

Enable inter-eye risk comparisons
Potentially enhance clinical efficiency

Methods (image grading)

3 graders, both eyes:

Drusen size and pigmentary abnormalities identified using retinal photography (Ferris et al. 2013)

Ferris, F.L. et al. (2013) ‘Clinical classification of age-related macular degeneration’, Ophthalmology, 120(4), pp. 844–851. Available at: https://doi.org/10.1016/j.ophtha.2012.10.036.

Ueda-Arakawa, N. et al. (2013) ‘Prevalence and genomic association of reticular pseudodrusen in age-related macular degeneration.’, American journal of ophthalmology, 155(2), pp. 260-269.e2. Available at: https://doi.org/10.1016/j.ajo.2012.08.011.

Posterior Eye Education & Research Society

15 of 51

Purpose and methods

Purpose

To develop an abridged risk model for AMD to:

Enable inter-eye risk comparisons
Potentially enhance clinical efficiency

Methods (image grading)

3 graders, both eyes:

Drusen size and pigmentary abnormalities identified using retinal photography (Ferris et al. 2013)
RPD identified using OCT (Ueda-Arakawa et al. 2013)

Ferris, F.L. et al. (2013) ‘Clinical classification of age-related macular degeneration’, Ophthalmology, 120(4), pp. 844–851. Available at: https://doi.org/10.1016/j.ophtha.2012.10.036.

Ueda-Arakawa, N. et al. (2013) ‘Prevalence and genomic association of reticular pseudodrusen in age-related macular degeneration.’, American journal of ophthalmology, 155(2), pp. 260-269.e2. Available at: https://doi.org/10.1016/j.ajo.2012.08.011.

Posterior Eye Education & Research Society

16 of 51

Purpose and methods

Purpose

To develop an abridged risk model for AMD to:

Enable inter-eye risk comparisons
Potentially enhance clinical efficiency

Methods (analysis)

Full person-level updated simplified AREDS applied

Outcomes = prognostic performance (AUC) and risk score separability (χ²)

Posterior Eye Education & Research Society

17 of 51

Purpose and methods

Purpose

To develop an abridged risk model for AMD to:

Enable inter-eye risk comparisons
Potentially enhance clinical efficiency

Methods (analysis)

Full person-level updated simplified AREDS applied

Outcomes = prognostic performance (AUC) and risk score separability (χ²)

One eye selected, and abridged eye-level risk models developed by:
Removing least predictive biomarkers (adjusted risks) sequentially 🡪 reduces #observations

Posterior Eye Education & Research Society

18 of 51

Purpose and methods

Purpose

To develop an abridged risk model for AMD to:

Enable inter-eye risk comparisons
Potentially enhance clinical efficiency

Methods (analysis)

Full person-level updated simplified AREDS applied

Outcomes = prognostic performance (AUC) and risk score separability (χ²)

One eye selected, and abridged eye-level risk models developed by:
Removing least predictive biomarkers (adjusted risks) sequentially 🡪 reduces #observations
Assigning approximate scores based on adjusted risk 🡪 reduces #risk scores

Posterior Eye Education & Research Society

19 of 51

Purpose and methods

Purpose

To develop an abridged risk model for AMD to:

Enable inter-eye risk comparisons
Potentially enhance clinical efficiency

Methods (analysis)

Full person-level updated simplified AREDS applied

Outcomes = prognostic performance (AUC) and risk score separability (χ²)

One eye selected, and abridged eye-level risk models developed by:
Removing least predictive biomarkers (adjusted risks) sequentially 🡪 reduces #observations
Assigning approximate scores based on adjusted risk 🡪 reduces #risk scores

Compared models

Posterior Eye Education & Research Society

20 of 51

Results

Study population

N = 269, averages ≈

Age 78 years
Follow-up 3.17 years
Sex 57% female
AMD severity 85% intermediate
RPD presence 50% in either eye

Posterior Eye Education & Research Society

21 of 51

Results

Study population

N = 269, averages ≈

Age 78 years
Follow-up 3.17 years
Sex 57% female
AMD severity 85% intermediate
RPD presence 50% in either eye

27% converted to late AMD over seven-years

Posterior Eye Education & Research Society

22 of 51

Results

Generating eye-level candidate risk models

Full model = updated simplified AREDS risk scale

Posterior Eye Education & Research Society

23 of 51

Results

Generating eye-level candidate risk models

Full model = updated simplified AREDS risk scale
Candidate #1 = remove 3 weakest biomarkers

Posterior Eye Education & Research Society

24 of 51

Results

Generating eye-level candidate risk models

Full model = updated simplified AREDS risk scale
Candidate #1 = remove 3 weakest biomarkers
Candidates #2-5 = from #1, further remove one more biomarker

Posterior Eye Education & Research Society

25 of 51

Results

Generating eye-level candidate risk models

Full model = updated simplified AREDS risk scale
Candidate #1 = remove 3 weakest biomarkers
Candidates #2-5 = from #1, further remove one more biomarker
Candidate #6 = from #1, redefine RPD to be eye-specific, i.e., remove RPD in the fellow eye

Posterior Eye Education & Research Society

26 of 51

Results

Prognostic performance of eye-level candidate risk models

Full model AUC = 76.2 ± 5.7 to 84.5 ± 5.9%

Posterior Eye Education & Research Society

27 of 51

Results

Prognostic performance of eye-level candidate risk models

Full model AUC = 76.2 ± 5.7 to 84.5 ± 5.9%

Candidates #2, #3, and #5 decreased prognostic performance

Posterior Eye Education & Research Society

28 of 51

Results

Prognostic performance of eye-level candidate risk models

Full model AUC = 76.2 ± 5.7 to 84.5 ± 5.9%

I.e., candidates #1, #4, and #6 were relatively unchanged

Posterior Eye Education & Research Society

29 of 51

Results

Separability of eye-level candidate risk scores

Full model χ²= significant overlap between most adjacent scores

Posterior Eye Education & Research Society

30 of 51

Results

Separability of eye-level candidate risk scores

Full model χ²= significant overlap between most adjacent scores

Candidates #1, #4, and #6 shown

Due to uncompromised prognostic performance

Posterior Eye Education & Research Society

31 of 51

Results

Separability of eye-level candidate risk scores

Full model χ²= significant overlap between most adjacent scores

Candidates #1, #4, and #6 shown

Due to uncompromised prognostic performance

Biomarkers assigned score of ‘1’ each

Due to similar adjusted risks in this population

Posterior Eye Education & Research Society

32 of 51

Results

Separability of eye-level candidate risk scores

Full model χ²= significant overlap between most adjacent scores

Candidates #1, #4, and #6 shown

Due to uncompromised prognostic performance

Biomarkers assigned score of ‘1’ each

Due to similar adjusted risks in this population

Candidate #4 showed significant overlaps

Posterior Eye Education & Research Society

33 of 51

Results

Separability of eye-level candidate risk scores

Full model χ²= significant overlap between most adjacent scores

Candidates #1, #4, and #6 shown

Due to uncompromised prognostic performance

Biomarkers assigned score of ‘1’ each

Due to similar adjusted risks in this population

Candidate #4 showed significant overlaps
I.e., candidates #1 and #6 showed separable risk scores

Posterior Eye Education & Research Society

34 of 51

Results

Separability of eye-level candidate risk scores

Full model χ²= significant overlap between most adjacent scores

Candidates #1, #4, and #6 shown

Due to uncompromised prognostic performance

Biomarkers assigned score of ‘1’ each

Due to similar adjusted risks in this population

Candidate #4 showed significant overlaps
I.e., candidates #1 and #6 showed separable risk scores
Candidate #6 was simpler

Posterior Eye Education & Research Society

35 of 51

Results summary

Clinical standard for predicting late AMD* = Ferris et al. 2005, AREDS report #18

Predicts progression to late AMD by person

2 key biomarkers (drusen size and pigmentary abnormalities)

Up to 7 total observations
5 annual risk scores (0-to-4)

*According to the American Academy of Ophthalmology Preferred Practice Pattern for AMD

Ferris, F.L. et al. (2005) ‘A simplified severity scale for age-related macular degeneration: AREDS report no. 18’, Archives of Ophthalmology, 123(11), pp. 1570–1574. Available at: https://doi.org/10.1001/archopht.123.11.1570.

Agrón, E. et al. (2024) ‘An updated simplified severity scale for age-related macular degeneration incorporating reticular pseudodrusen: Age-Related Eye Disease Study report number 42’, Ophthalmology, 131(10), pp. 1164–1174. Available at: https://doi.org/10.1016/j.ophtha.2024.04.011.

Posterior Eye Education & Research Society

36 of 51

Results summary

Clinical standard for predicting late AMD* = Ferris et al. 2005, AREDS report #18

Predicts progression to late AMD by person

2 key biomarkers (drusen size and pigmentary abnormalities)

Up to 7 total observations
5 annual risk scores (0-to-4)

Updated clinical standard for predicting late AMD* = Agrón et al. 2024, AREDS report #42

Predicts progression to late AMD by person

3 key biomarkers (drusen size, pigmentary abnormalities, RPD)

Up to 9 total observations
10 annual risk scores (0-to-4, without and with RPD)

*According to the American Academy of Ophthalmology Preferred Practice Pattern for AMD

Ferris, F.L. et al. (2005) ‘A simplified severity scale for age-related macular degeneration: AREDS report no. 18’, Archives of Ophthalmology, 123(11), pp. 1570–1574. Available at: https://doi.org/10.1001/archopht.123.11.1570.

Agrón, E. et al. (2024) ‘An updated simplified severity scale for age-related macular degeneration incorporating reticular pseudodrusen: Age-Related Eye Disease Study report number 42’, Ophthalmology, 131(10), pp. 1164–1174. Available at: https://doi.org/10.1016/j.ophtha.2024.04.011.

Posterior Eye Education & Research Society

37 of 51

Results summary

Clinical standard for predicting late AMD* = Ferris et al. 2005, AREDS report #18

Predicts progression to late AMD by person

2 key biomarkers (drusen size and pigmentary abnormalities)

Up to 7 total observations
5 annual risk scores (0-to-4)

Updated clinical standard for predicting late AMD* = Agrón et al. 2024, AREDS report #42

Predicts progression to late AMD by person

3 key biomarkers (drusen size, pigmentary abnormalities, RPD)

Up to 9 total observations
10 annual risk scores (0-to-4, without and with RPD)

Possible options for predicting late AMD = Trinh et al. 202X

Predicts progression to late AMD by eye

3 key biomarkers (drusen size, pigmentary abnormalities, RPD)

Up to 4 total observations
5 annual risk scores (0-to-4)

*According to the American Academy of Ophthalmology Preferred Practice Pattern for AMD

Ferris, F.L. et al. (2005) ‘A simplified severity scale for age-related macular degeneration: AREDS report no. 18’, Archives of Ophthalmology, 123(11), pp. 1570–1574. Available at: https://doi.org/10.1001/archopht.123.11.1570.

Agrón, E. et al. (2024) ‘An updated simplified severity scale for age-related macular degeneration incorporating reticular pseudodrusen: Age-Related Eye Disease Study report number 42’, Ophthalmology, 131(10), pp. 1164–1174. Available at: https://doi.org/10.1016/j.ophtha.2024.04.011.

Posterior Eye Education & Research Society

38 of 51

Take-away message

Possible options for predicting late AMD = Trinh et al. 202X

Predicts progression to late AMD by eye

3 key biomarkers (drusen size, pigmentary abnormalities, RPD)

Up to 4 total observations
5 annual risk scores (0-to-4)

Example 3-year prediction

Posterior Eye Education & Research Society

39 of 51

Discussion

We provide options for tweaking the updated simplified AREDS risk model

By reducing the number of observations, i.e., turning it into an eye-level (rather than person-level) prediction
By reducing the number of risk scores, i.e., re-configuring the scoring system

Posterior Eye Education & Research Society

40 of 51

Discussion

We provide options for tweaking the updated simplified AREDS risk model

By reducing the number of observations, i.e., turning it into an eye-level (rather than person-level) prediction
By reducing the number of risk scores, i.e., re-configuring the scoring system

When is an eye-specific prediction useful?

We don’t manage pxs by eye…

Posterior Eye Education & Research Society

41 of 51

Discussion

We provide options for tweaking the updated simplified AREDS risk model

By reducing the number of observations, i.e., turning it into an eye-level (rather than person-level) prediction
By reducing the number of risk scores, i.e., re-configuring the scoring system

When is an eye-specific prediction useful?

We don’t manage pxs by eye…except:

If px already has late AMD in one eye

Instead of checking for RPD in either eye, can use the simpler scale and check in one eye

Trivizki, O. et al. (2023) ‘Symmetry of macular fundus features in age-related macular degeneration’, Ophthalmology Retina, 7(8), pp. 672–682. Available at: https://doi.org/10.1016/j.oret.2023.03.016.

Posterior Eye Education & Research Society

42 of 51

Discussion

We provide options for tweaking the updated simplified AREDS risk model

By reducing the number of observations, i.e., turning it into an eye-level (rather than person-level) prediction
By reducing the number of risk scores, i.e., re-configuring the scoring system

When is an eye-specific prediction useful?

We don’t manage pxs by eye…except:

If px already has late AMD in one eye

Instead of checking for RPD in either eye, can use the simpler scale and check in one eye

If px already has other severe vision loss in one eye

Trivizki, O. et al. (2023) ‘Symmetry of macular fundus features in age-related macular degeneration’, Ophthalmology Retina, 7(8), pp. 672–682. Available at: https://doi.org/10.1016/j.oret.2023.03.016.

Posterior Eye Education & Research Society

43 of 51

Discussion

We provide options for tweaking the updated simplified AREDS risk model

By reducing the number of observations, i.e., turning it into an eye-level (rather than person-level) prediction
By reducing the number of risk scores, i.e., re-configuring the scoring system

When is an eye-specific prediction useful?

We don’t manage pxs by eye…except:

If px already has late AMD in one eye

Instead of checking for RPD in either eye, can use the simpler scale and check in one eye

If px already has other severe vision loss in one eye
If px has considerable AMD asymmetry

Half of cases are asymmetric (Trivizki et al. 2023)

Trivizki, O. et al. (2023) ‘Symmetry of macular fundus features in age-related macular degeneration’, Ophthalmology Retina, 7(8), pp. 672–682. Available at: https://doi.org/10.1016/j.oret.2023.03.016.

Posterior Eye Education & Research Society

44 of 51

Discussion

We provide options for tweaking the updated simplified AREDS risk model

By reducing the number of observations, i.e., turning it into an eye-level (rather than person-level) prediction
By reducing the number of risk scores, i.e., re-configuring the scoring system

When is an eye-specific prediction useful?

We don’t manage pxs by eye…except:

Sometimes there may be one particular eye of interest in clinic

Trivizki, O. et al. (2023) ‘Symmetry of macular fundus features in age-related macular degeneration’, Ophthalmology Retina, 7(8), pp. 672–682. Available at: https://doi.org/10.1016/j.oret.2023.03.016.

Posterior Eye Education & Research Society

45 of 51

Discussion

We provide options for tweaking the updated simplified AREDS risk model

By reducing the number of observations, i.e., turning it into an eye-level (rather than person-level) prediction
By reducing the number of risk scores, i.e., re-configuring the scoring system

When is an eye-specific prediction useful?

We don’t manage pxs by eye…except:

Sometimes there may be one particular eye of interest in clinic
For research study designs

If only one eye available, e.g., missing or poor quality fellow eye data

Trivizki, O. et al. (2023) ‘Symmetry of macular fundus features in age-related macular degeneration’, Ophthalmology Retina, 7(8), pp. 672–682. Available at: https://doi.org/10.1016/j.oret.2023.03.016.

Posterior Eye Education & Research Society

46 of 51

Discussion

We provide options for tweaking the updated simplified AREDS risk model

By reducing the number of observations, i.e., turning it into an eye-level (rather than person-level) prediction
By reducing the number of risk scores, i.e., re-configuring the scoring system

When is an eye-specific prediction useful?

We don’t manage pxs by eye…except:

Sometimes there may be one particular eye of interest in clinic
For research study designs

If only one eye available, e.g., missing or poor quality fellow eye data
May wish to randomise interventions at the eye-level, i.e., allocating one eye tx, one eye control

Trivizki, O. et al. (2023) ‘Symmetry of macular fundus features in age-related macular degeneration’, Ophthalmology Retina, 7(8), pp. 672–682. Available at: https://doi.org/10.1016/j.oret.2023.03.016.

Posterior Eye Education & Research Society

47 of 51

Discussion

We provide options for tweaking the updated simplified AREDS risk model

By reducing the number of observations, i.e., turning it into an eye-level (rather than person-level) prediction
By reducing the number of risk scores, i.e., re-configuring the scoring system

When is an eye-specific prediction useful?

We don’t manage pxs by eye…except:

Sometimes there may be one particular eye of interest in clinic
For research study designs

If only one eye available, e.g., missing or poor quality fellow eye data
May wish to randomise interventions at the eye-level, i.e., allocating one eye tx, one eye control
May wish to simplify pseudo-random matching (e.g., propensity-score matching) using a composite eye-level risk% instead of individual biomarkers, to preserve sample size

Trivizki, O. et al. (2023) ‘Symmetry of macular fundus features in age-related macular degeneration’, Ophthalmology Retina, 7(8), pp. 672–682. Available at: https://doi.org/10.1016/j.oret.2023.03.016.

Posterior Eye Education & Research Society

48 of 51

Discussion

We provide options for tweaking the updated simplified AREDS risk model

By reducing the number of observations, i.e., turning it into an eye-level (rather than person-level) prediction
By reducing the number of risk scores, i.e., re-configuring the scoring system

When is an eye-specific prediction useful?

We don’t manage pxs by eye…except:

Sometimes there may be one particular eye of interest in clinic
For research study designs, to account for inter-eye risks

Posterior Eye Education & Research Society

49 of 51

Limitations and future directions

Small, contained sample
Will need validation in much larger, longer, and less biased cohorts

Posterior Eye Education & Research Society

50 of 51

Limitations and future directions

Small, contained sample
Will need validation in much larger, longer, and less biased cohorts

Impact of different risk scales on clinical practice
Gaze-tracking analyses
Decision curve analyses, i.e., influence on referral patterns and patient outcomes

Posterior Eye Education & Research Society

51 of 51

Thank you for your attention

Feel free to email me any (nice) questions/thoughts/criticisms: m.trinh@unsw.edu.au

Possible options for predicting late AMD = Trinh et al. 202X

Predicts progression to late AMD by eye

3 key biomarkers (drusen size, pigmentary abnormalities, RPD)

Up to 4 total observations
5 annual risk scores (0-to-4)

Posterior Eye Education & Research Society