**Purpose**:
To determine the relationship between the progression of geographic atrophy (GA) and its baseline area (BA) using the square root transformation (sqrt) for different atrophy sizes.

**Methods**:
Single eyes of patients with GA visiting the Institut de la Màcula (Barcelona, Spain) between December 2009 and January 2018 with a follow-up of ≥6 months were included. The main outcome was the correlation between BA and growth after the sqrt using Pearson's *r* and Spearman's *rho*. The graphical relationship was explored using linear and LOWESS regression. In a secondary, prespecified analysis, progression was compared by BA categories (Age-Related Eye Disease Study [AREDS] classification and BA tertiles). In post hoc analyses, the results were evaluated in subgroups defined by location of atrophy, number of lesions, fundus autofluorescence pattern, and fellow-eye status.

**Results**:
We included 128 eyes (mean follow-up, 3.1 years). The correlation between BA and progression was negative (*r* = −0.30, *P* = 0.0005; *rho* = −0.25, *P* = 0.0042). There was a decrease in the rate of progression in mm/year with increasing BA, but this was significant for tertiles (*P* = 0.0078) and not AREDS (*P* = 0.20). The descending trend was driven by high-risk features.

**Conclusions**:
The correlation between GA progression and BA using the sqrt is negative. This has implications for the expected prediction of progression of a given lesion and to avoid overestimating the beneficial effects of interventional therapies.

**Translational Relevance**:
The GA progression/BA relationship using the sqrt currently is regarded as independent. Our results suggest the sqrt slope actually is negative, which should be kept in mind to avoid misinterpretation of results in advanced therapies.

^{1}Progression of atrophy is negatively related to visual function and everyday task activities,

^{2–5}and it is now accepted as an endpoint in clinical trials.

^{6}

^{7–9}To decrease test–retest variability and avoid the dependence of progression rate on baseline lesion size, Yehoshua et al.

^{10}proposed taking the square root of the area to measure GA progression: the square root transformation (sqrt). This measured lineal (mm/year) rather than surface (mm

^{2}/year) progression and removed the correlation between BA and progression in many studies, which permitted to focus in other factors driving progression.

^{11–13}

^{2,8,11,14,15}Therefore, the relationship with progression rate may differ according to baseline lesion size even if measured with sqrt. This may be relevant for treatments where few patients with large lesions typically are included, such as stem-cell therapies. In these studies, it is common to use a paired design with the worst eye being selected systematically for the active group; if baseline size differs between eyes, an incorrect assumption about the treatment efficacy could be made.

^{9}was a prospective natural history study of factors associated with GA progression conducted from December 2009 until August 2013; the medical history and imaging of these patients up to January 2018 were reviewed to increase their long term follow-up. To increase the sample size, we included new patients who met eligibility criteria who visited the clinic up to January 2018. The study followed the tenets of the Declaration of Helsinki and was approved by the Centro Médico Teknon ethics committee. An informed consent was obtained from all individual participants included in the study after explanation of the nature and possible consequences of the study.

^{16}secondary to AMD and who had a follow-up ≥6 months were potentially eligible. This new definition of GA is less restrictive than that used previously (an area of RPE atrophy >0.5 disk areas [approximately 1.27 mm

^{2}]) and allowed the inclusion of more patients.

^{9}Rate of progression was determined by subtracting area of atrophy (in mm

^{2}) at the last visit from area of atrophy at baseline divided by time between visits (in years). For the sqrt measurements of progression (mm/year), the square root of the area on the last and first visits was calculated, subtracted, and the result was again divided by time between visits (in years).

*r*and Spearman

*rho*correlation coefficients. This relationship also was plotted using linear regression and locally weighted scatterplot smoothing (LOWESS) curves with a tricube weighted function. These analyses were repeated in the original scale (mm

^{2}and mm

^{2}/year) to check if the relationship was as expected.

^{11}and by tertiles of BA using 1-way analysis of variance (ANOVA) tests. As a post hoc analysis, the relationship also was tested across subgroups defined by location of atrophy (foveal versus extrafoveal), number of lesions (unifocal versus multifocal), FAF pattern (“none” and “focal” versus “banded” and “diffuse”),

^{17}and diagnosis of the fellow eye (drusen versus late AMD, GA or neovascular AMD).

*P*value <0.05 was considered statistically significant.

^{2}, mean sqrt at baseline was 2.44 (1.16) mm, and mean progression was 1.86 (1.12) mm

^{2}/year, with a sqrt of 0.33 (0.21) mm/year. Regarding lesion characteristics, 41.4% (53/128) had foveal atrophy; 67.2% (86/128) were multifocal; FAF pattern was none or focal in 29.7% (38/128), banded or diffuse in 67.2% (86/128), and other in 3.1% (4/128); and diagnosis of the fellow eye was drusen in 15.6% (20/128), late AMD in 80.5% (103/128), and other in 3.9% (5/128).

*r*= −0.30,

*P*= 0.0005 and Spearman's

*rho*= −0.25,

*P*= 0.0042. As expected, the correlation was positive for progression expressed in mm

^{2}/year, with Pearson's

*r*and Spearman's

*rho*= 0.34,

*P*= 0.0001. These results are shown in Figure 1.

^{2}/year and mm/year, using different classifications for BA (in mm

^{2}): AREDS and tertiles of the current sample. The results are shown in the Table, and were statistically significant for the sqrt tertiles of BA in our study (

*P*= 0.0078), but not for the AREDS classification (

*P*= 0.20). However, if the small lesion size category (0.5 to <0.75 DA) was increased to include lesions <0.5 DA (

*n*= 27), then the progression in sqrt in this category became 0.44 (0.35) mm/year, and the comparison between categories reached statistical significance (

*P*= 0.01). Therefore, the decreasing trend also was observed for BA in mm

^{2}. As expected, for measurements in mm

^{2}/year progression rate increased with increasing BA regardless of the classification used to stratify baseline lesion size (

*P*≤ 0.0002).

*r*between −0.33 and −0.42 and

*rho*between −0.27 and −0.40 (all

*P*≤ 0.01). This was not observed for lower-risk categories (foveal atrophy, unifocal lesions, patterns with no/minimally increased FAF, and fellow eyes with drusen), with

*r*between +0.09 and −0.20 and

*rho*between +0.13 and −0.12 (all

*P*≥ 0.16).

^{2}(Table and Supplementary Figure S1).

*r*= −0.09, Spearman's

*rho*= −0.06),

^{10}but this did not reach statistical significance (

*P*> 0.40). This finding supported the assumption that the dependence on BA was eliminated with the sqrt. An increased magnitude of the correlation, a larger sample size, and/or a wider range of lesion sizes in our study may explain why some of our results were statistically significant, while others did not reach this conclusion.

^{11,14,15}A decrease in perimeter translates into less number of diseased RPE cells in contact with adjacent healthier cells, which may slow the lateral spread of the atrophy.

^{14}

^{18}As for the group overall, we found a moderate negative correlation in eyes with high-risk characteristics across the four groups, while eyes with low-risk characteristics showed a very low correlation. This suggested that growth in this metric slows down in very large lesion sizes in these particular subgroups. Small sample size in some categories with low-risk characteristics (i.e., in fellow eye diagnosis there were just 20 eyes with drusen) make results susceptible to a few outliers. Also, large lesions at baseline are required to observe this descending trend: we have shown that as lesions grow the linear progression (mm/year) decreases, probably because the growth is distributed along the whole perimeter of an increasingly large atrophic lesion. Figure 2 shows that the number of eyes with low-risk characteristics with a large BA square root (≥4 mm) is low (location of atrophy, multifocal lesions) or zero (FAF pattern, fellow eye), which may preclude the observation of this phenomenon. In fact, eyes may change from one category to another as BA enlarges. It is uncommon to see drusen only in the fellow eye of a patient with very large GA, but it is not rare to see them in the fellow eye of a patient with a small GA lesion. Patterns none and focal on FAF also are rarely observed in very large lesions.

^{17,19}Given the exploratory nature of these analyses, they are regarded as hypothesis-generating and should be confirmed in further studies.

^{10}Nonetheless, it still may be reasonable to adjust for BA when evaluating risk factors for GA progression or when testing the efficacy of new therapies, especially in uncontrolled or small trials were randomization may not suffice to achieve a perfect balance between study arms in terms of important predictors. Otherwise, confounding may creep in. Special care should be taken in the context of advanced therapies, particularly stem cell treatments. In these studies, a paired design is not uncommon and the worst eye (usually that with the larger atrophy) may be selected systematically for treatment. If the sqrt is used, a slower progression is to be expected in the treated eye even in the absence of a real treatment effect, and, thus, claims of therapy efficacy may be mistakenly raised (Fig. 3).

**Monés J.**, None;

**Biarnés M.**, None

*Invest Ophthalmol Vis Sci*. 2009; 50: 4982–4991.

*Ophthalmology*. 1999; 106: 1768–1779.

*Ophthalmology*. 1996; 103: 1458–1466.

*Invest Ophthalmol Vis Sci*. 2011; 52: 1119–1126.

*Ophthalmology*. 2018; 125: 842–849.

*Retina*. 2016; 36: 1806–1822.

*Ophthalmology*. 2007; 114: 271–277.

*Ophthalmology*. 2016; 123: 361–368.

*Am J Ophthalmol*. 2015; 160: 345–353.

*Ophthalmology*. 2011; 118: 679–686.

*JAMA Ophthalmol*. 2013; 131: 110.

*Ophthalmology*. 2014; 121: 693–701.

*Ophthalmology*. 2015; 122: 1356–1365.

*Ophthalmology*. 2013; 120: 2666–2671.

*Retina*. https://doi.org/10.1097/IAE.0000000000002206.

*Ophthalmology*. 2018; 125: 537–548

*Am J Ophthalmol*. 2007; 143: 463–472.

*Ophthalmology*. 2018; 125: 369–390.

*Invest Ophthalmol Vis Sci*. 2014; 55: 8302–8308.