**Purpose**:
This study aims to determine whether OCT-derived rates of change in minimum rim width (MRW) are associated with and can potentially predict corresponding alterations in retinal nerve fiber layer thickness (RNFLT) in people with glaucoma.

**Methods**:
The rates of change between six-monthly visits were taken from 568 eyes of 278 participants in the P3 Study. Structural equation models (SEM) assessed whether one parameter was predicted by the concurrent or previous rate of the other parameter, after adjusting for its own rate in the previous time interval. Root mean square error of approximation (RMSEA, with 90% confidence intervals [CI]), Tucker Lewis index (TLI) and the comparative fit index (CFI) assessed goodness of fit.

**Results**:
Models without a time lag provided a better fit for the data (RMSEA = 0.101 [CI, 0.089, 0.113]), compared to a model featuring a time lag in RNFLT (RMSEA = 0.114 [CI, 0.102, 0.126]) or MRW (RMSEA = 0.114 [CI, 0.102, 0.127]). The SEMs indicated that rates for both MRW and RNFLT were predicted by their own rate in the previous time interval and by the other measure's change in the concurrent time interval (*P* > 0.001 for all). No evidence of a clinically significant time lag for either parameter was determined.

**Conclusions**:
MRW and RNFLT exhibit concurrent changes over time in patients with glaucoma, with no clinically significant time lag determined.

**Translational Relevance**:
RNFLT may be more useful than MRW in early glaucoma assessment because of its previously reported lower variability and reduced sensitivity to intraocular pressure changes.

^{1}While functional tests such as VF perimetry have traditionally served as the cornerstone for assessing glaucoma progression, measurement of the structural changes remains key due to better repeatability, objectivity, and quicker acquisition time. Optical coherence tomography (OCT), a widely used tool, is used to monitor glaucomatous changes such as alterations in RNFL thickness (RNFLT), contributing crucial insights into the structural aspect of glaucoma. Indeed, RNFLT has been demonstrated to be directly related to the number of retinal ganglion cell axons that remain.

^{2}

^{3}

^{,}

^{4}

^{5}Increased cupping exerts a mechanical influence on the axons, which may result in a reduction of MRW.

^{6}Importantly, RNFLT measurement occurs sufficiently outside the cup (typically along a 6° radius circle centered on the BMO centroid), minimizing its susceptibility to these mechanical effects.

^{7}

^{,}

^{8}This is postulated to be the reason why MRW tends to exhibit greater variability than RNFLT when IOP fluctuates.

^{9}

^{–}

^{11}In this context, the reduction in MRW due to increased cupping may potentially precede and predict actual axon loss and, therefore, RNFL thinning. Notably, in a nonhuman primate experimental glaucoma model, changes in MRW precede the onset of RNFLT change.

^{12}MRW has also been reported to show earlier detectable change than RNFLT in early-stage glaucoma,

^{3}and the rate of change of MRW has been found to be consistently greater than RNFLT in patients with early normal-tension glaucoma (NTG).

^{13}However, Shi et al.

^{14}reported that RNFLT is more likely than MRW to reveal a declining trend over time in patients with central or moderate-to-advanced glaucomatous damage. The potential time lag between MRW and RNFLT alterations underscores the complexity of glaucomatous progression and highlights the need for comprehensive assessments that consider the distinct dynamics of these parameters.

^{15}Automated layer segmentations were manually corrected by technicians if necessary.

^{16}All included OCT scans had a quality score ≥16; quality scores below this were counted as missing data for that visit. Data were only retained if both RNFLT and MRW values were available from scans of high quality conducted in the same session. In addition to OCT imaging, all participants underwent SAP using a Humphrey Field Analyzer II (Carl Zeiss Meditec Inc., Dublin, CA, USA), with the SITA Standard testing strategy and 24-2 test pattern, on the same day. The Glaucoma hemifield test was used for automated evaluation of localized visual field loss occurrences.

^{17}

*MRWRate*=

_{n}*(MRW*. The main analysis was conducted on a time series of four rates, using data from five visits, therefore the duration of the data analyzed was approximately two to three years. A secondary analysis was performed using longer time series of 10 rates using data from 11 visits. For this subanalysis the duration of the data analyzed was approximately five to six years.

_{n+1}–MRW_{n})/(Date_{n+1}–Date_{n})*MRW*was unavailable, then

_{n+1}*MRWRate*was treated as missing data, and

_{n}*MRWRate*=

_{n+1}*(MRW*. This was extended to accommodate up to three consecutive time periods with missing visits.

_{n+2}– MRW_{n})/(Date_{n+2}– Date_{n})^{18}SEM is well suited for analyzing changes over time because it can accommodate repeated measures of variables and assess how variables change in relation to one another across multiple time points. Furthermore, SEM provides fit indexes to assess how well the proposed model fits the observed data. Researchers can modify and refine their models based on fit indexes, ensuring that the model adequately represents the underlying data structure. SEM has been used previously by Gardiner et al.

^{19}to demonstrate that changes in SAP precede and predict changes in RNFLT.

*MRWRate*(i.e., the “current” rate, denoted by “

_{n}_{n}”) and

*MRWRate*(the previous rate, denoted by “

_{n-1}_{n-}

_{1}”)

*;*and

*RNFLTRate*and

_{n}*RNFLTRate*, respectively. We will test four SEM models to make these comparisons, which we will refer to as Models A to D.

_{n-1}*ΔMRWRate*(i.e., from visit n − 1 to visit n) will be inversely correlated with the

_{n-1}*ΔMRWRate*(i.e., from visit n to visit n + 1), because they both have the measurement at visit n in common. We assume that the true rate of axon loss is approximately linear; this is reasonable as the time series are relatively short (five visits, which equates to approximately three years). Figure 1 shows the path diagram for Model A.

_{n}_{A}is significant, knowing the rate of change of RNFLT improves predictions of the concurrent rate of change of MRW. If β

_{B}is significant, the rate of change of RNFLT over the previous time period helps predict the rate of MRW change in the current period; that is, there is a time lag with RNFLT changing earlier than MRW. If β

_{C}is significant, knowing the rate of change of MRW improves predictions of the current rate of change of RNFLT. Last, if β

_{D}is significant, the rate of change of MRW over the previous time period helps predict the rate of RNFLT change in the current period; that is, there is a time lag with MRW changing earlier than MRW.

^{20}Models were fit using the maximum likelihood estimation because the estimates are unbiased, have the smallest possible variances, and are asymptotically normally distributed, despite the presence of data missingness. Goodness of fit for each model was assessed using the root mean square error of approximation (RMSEA, chosen as our primary measure of goodness of fit because it is an absolute fit index and gives 90% confidence intervals [CI]),

^{21}as well as the Tucker Lewis index (TLI)

^{22}and the comparative fit index (CFI), both examples of incremental fit indexes.

^{23}For context, RMSEA values closer to zero and CFI and TLI values closer to one represent a good fit. A correlational analysis was performed using Spearman's rho, and a Steiger's

*Z*-test was used to evaluate differences between two dependent correlation coefficients.

^{17}and OCT structural values within the normative limits

^{24}and (ii) eyes rated structurally or functionally “abnormal,” who fell outside the above criteria. To assess the influence of age on ONH deformation, we also assessed a subgroup of patients: (i) aged <65 years and (ii) aged ≥65 years. The SEM-based analysis was repeated in each of these cohorts.

*ΔMRWRate*and

_{n-1}*ΔRNFLTRate*were significant predictors of

_{n}*ΔMRWRate*(both

_{n}*P*< 0.001), which is to say that the previous rate of MRW and the concurrent rate of RNFLT significantly predict the current rate of MRW. We did not find evidence that

*ΔRNFLTRate*significantly predicts

_{n-1}*ΔMRWRate*. Therefore no evidence of a clinically significant time lag for RNFLT parameter was determined. The coefficient estimates and the respective significance levels for all four models can be found in Figure 2. The coefficient estimates and intercepts that best fit Model A are as follows:

_{n}*ΔRNFLTRate*and

_{n-1}*ΔMRWRate*were significant predictors of

_{n}*ΔRNFLTRate*(both

_{n}*P*≤ 0.001), which as before means that the previous rate of RNFLT and the concurrent rate of MRW significantly predicts the current rate of RNFLT. We did not find evidence that

*ΔMRWRate*significantly predicts

_{n-1}*ΔRNFLTRate*. Hence, no evidence of a clinically significant time lag for MRW parameter was determined. The coefficient estimates and intercepts that best fit Model C are as follows:

_{n}^{3}), we repeated the SEMs using data from eyes rated functionally or structurally within normal limits. Still, no clinically significant time lag for either MRW nor RNFLT was identified. This data was compared to our remaining cohort of patients outside these normal limits, and no significant difference in model coefficients was identified.

^{6}As the RNFLT parameter we measured is further away from this region (6° from the BMO centroid), and thus is less influenced by direct effects of IOP,

^{7}

^{,}

^{8}it follows that changes in MRW would be expected to occur

*before*changes in RNFLT. Indeed, He and colleagues identified changes in MRW preceded that of RNFLT in an experimental glaucoma model, therefore concluding a time lag in MRW to be evident in nonhuman primates.

^{12}Furthermore, Chauhan et al.

^{3}reported that MRW was found to have earlier detectable change compared to RNFLT parameters in early-stage glaucoma.

^{25}

^{,}

^{26}Consequently, there is less susceptibility to mechanical deformation at the ONH.

^{27}This may result in MRW thinning concurrently to RNFLT as seen in this study or that the structural damage for the two parameters has a similar onset.

^{28}and cannot be feasibly controlled for in human experimental studies.

^{28}

^{,}

^{29}Therefore any IOP measures or indeed parameters impacted by IOP changes such as MRW, are a mere snapshot of these fluctuating changes and may not reflect the more chronic structural glaucomatous changes in the eye. Conversely, in experimental glaucoma studies, it is possible (and standard practice) to stabilize IOP to reliably assess OCT metrics. Thus this difference in experimental procedure may explain why He and colleagues

^{12}found a time lag, whereas this study did not. Further elaborating on the impact of transient cupping as a result of IOP changes, not only is it known that fluctuating IOP and susceptibility to IOP-related damage varies across individuals,

^{30}but the cohort in the P3 study are managed clinically and are therefore in receipt of medication aimed to reduce IOP. Thus the MRW parameters in this study can be expected to be more variable, potentially obscuring any true time lag.

^{11}

^{31}so such a time lag would be too short to be relevant clinically.

^{3}

^{,}

^{12}In this scenario, the first stage would consist of both conformational and remodeling-based changes at the ONH that alter MRW but not RNFLT; followed by a second stage at which axon loss occurs and both MRW and RNFLT change concurrently. Our results then suggest that the transition between those stages occurs relatively abruptly, rather than a gradual transition over several years, which would be detectable as a time lag in our models.

^{32}In light of this consideration, future research should explore lamina cribrosa displacement to achieve a more comprehensive understanding of optic nerve head deformation in the context of glaucoma. Last, it is important to note that factors potentially influencing the rate of change in MRW, including IOP and the use of ocular hypertensive medications, were not explored as covariates in our study. However, preliminary analysis did not reveal a significant correlation between IOP and the rate of change in MRW and RNFLT in this treated cohort. Additionally, we assume that the primary effect of ocular hypertensive medication would be to lower IOP, contingent on patient adherence and, hence, that they are unlikely to affect the time lag between different measurements. Due to these considerations, incorporating these factors as covariates in our models was deemed unfeasible.

**B.E. Higgins**, None;

**H. Yang**, None;

**S.K. Gardiner**, Heidelberg Engineering (F, R)

*JAMA*. 2014; 311: 1901. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2016; 57: OCT404–OCT412. [CrossRef] [PubMed]

*Ophthalmology*. 2013; 120: 535–543. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2012; 53: 1852. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2011; 52: 7109. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2019; 60: 26–35. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2011; 52: 1206. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2009; 50: 4719. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2011; 52: 9431–9437. [CrossRef] [PubMed]

*Ophthalmology*2011; 118: 52–59. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2015; 56: 6886–6891. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2014; 55: 574. [CrossRef] [PubMed]

*J Clin Med*. 2020; 9: 1–13.

*Am J Ophthalmol*. 2023; 253: 181–188. [CrossRef] [PubMed]

*Am J Ophthalmol*. 2014; 157: 540–549.e2. [CrossRef] [PubMed]

*Am J Ophthalmol*. 2017; 174: 1–8. [CrossRef] [PubMed]

*Arch Ophthalmol*. 1992; 110: 812–819. [CrossRef] [PubMed]

*Structural Equation Modeling (2nd ed): Foundations and Extensions*. New York: Sage Publications; 2009.

*Invest Ophthalmol Vis Sci*. 2020; 61: 5–5. [CrossRef] [PubMed]

*J Stat Softw*. 2012; 48: 1–36. [CrossRef]

*Methods Psychol Res*. 2003; 8: 23–74.

*Psychometrika*. 1973; 38: 1–10. [CrossRef]

*Psychol Bull*. 1990; 107: 238–246. [CrossRef] [PubMed]

*Am J Ophthalmol*. 2020; 213: 203. [CrossRef] [PubMed]

*J Mech Behav Biomed Mater*. 2012; 16: 181–191. [CrossRef] [PubMed]

*Br J Ophthalmol*. 2000; 84: 318–323. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2014; 55: 1531–1539. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2011; 52: 1896. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2019; 60: 2572. [CrossRef] [PubMed]

*Exp Eye Res*. 2015; 133: 19. [CrossRef] [PubMed]

*Ophthalmology*. 2021; 128: P151–P192. [CrossRef] [PubMed]

*Ophthalmology*. 2015; 122: 1317–1329. [CrossRef] [PubMed]