The diagnosis of glaucoma involves a set of characteristic optic nerve findings usually accompanied by VF changes. It is known that pathologic changes of the optic nerve precede VF changes as measured with standard achromatic perimetry to such an extent that 25% to 35% of axons can be lost at a given location before the visual threshold ventures outside of the normal range.3–5
This observation has stimulated interest in measuring RNFL thickness as a potential method of recognizing axon loss in advance of recognizing VF abnormality. Indeed, clinical grading of RNFL is highly predictive of future VF defect development.8
Scanning laser polarimetry and optical coherence tomography are 2 methods currently available for quantitatively assessing the RNFL.
The current study demonstrates that OCT measures the RNFL as thinner in older persons, with a decline of approximately 2 μ
m per decade. Because there have been no OCT longitudinal studies of RNFL thickness in healthy individuals, we make the assumption that a cross-sectional analysis is a good surrogate for change over time in an individual. Cross-sectional studies of RNFL thickness using scanning laser polarimetry also have found a decrease in RNFL thickness with age.9–12
The human RNFL loses approximately 5000 axons per year from birth to death, approximately 2500 per year before age 50 and 7500 per year after 50.3
It is not surprising, then, that RNFL thickness decreases with age. Our findings indicate a loss per year of 0.2 μ
m, or 0.2% per year loss in the 100-μ
m mean thickness for older adults studied. The loss of 7500 axons from the total of about 1 000 000 axons in the normal adult is 0.75% per year. Because the loss of RNFL thickness and loss of optic nerve axons as a proportion of the total number are on the same order of magnitude, the 2 sets of data support the likelihood that there is a modest loss of RGCs with age. Likewise, the general similarity of the 2 estimates for age-related loss, clinical and histological, in a sense provides mutual validity for the 2 approaches. Several earlier studies have shown a decrease in RNFL thickness with age by OCT7,13
or in RNFL axons by histologic analysis.14,15
A third histologic study failed to find any difference in the number of axons between older and younger individuals.16
Conflicting results in histologic studies may be because these studies look at relatively few subjects and there is a large variation in the number of RNFL axons in normal individuals, between 700 000 and 1.4 million. Unless one looks at a large number of normal subjects, trends such as this might be missed.
The clinical implications of the finding that RNFL thickness measurements decreases with age cannot be overemphasized. This decline should be taken into consideration when interpreting the lower limits of the normal range for diagnosis. For instance, an RNFL thickness of 80 μm may be normal for a 70-year-old but would be abnormally low for the average 40-year-old. This age-related change in RNFL thickness is taken into account by the Stratus OCT software to avoid confusing the normal aging effect with glaucomatous RNFL thinning; however, at 2 μm per decade, the effect of aging on RNFL thickness, though statistically significant, is still clinically small.
Ethnicity was associated modestly with RNFL thickness in the current study, though in the regression model all ethnic designations had 95% confidence limits that overlapped 1. The present Stratus OCT normative database has relatively few subjects in the African American, Asian, and East Indian subgroups, as shown in , so conclusions about ethnicity must be interpreted with caution. Thus, further study may be worthwhile for those of Asian, African, or East Indian heritage. We did find that persons of Hispanic ethnicity had a significantly thicker mean RNFL than those of European derivation. Perhaps future studies will clarify whether ethnic differences in RNFL exist and need to be taken into account when diagnosing glaucoma using RNFL thickness measurements. In addition, ethnicity-specific databases may need to be developed for accurate use of the normative database software.
The current study found that RNFL thickness was related significantly to both axial length and refractive error. Longer eyes and more myopic eyes had a thinner measured RNFL. A recent study by Hoh et al17
failed to demonstrate a relationship between peripapillary RNFL thickness and refractive error or axial length in 132 young Asian male military subjects. Limitations of that study included the use of first-generation technology OCT; a relatively small sample size; lack of gender, age, or ethnic diversity; and the use of univariate analysis. Also, a small study using scanning laser polarimetry in 43 normal subjects failed to find a relationship between refractive error or axial length and RNFL thickness.18
Because the current study measured all eyes at a fixed angular distance from the geometric center of the disc, several variables could play a role in how variations in ocular anatomy impact RNFL thickness as determined by OCT. If all eyes had the same number of RGCs, with one axon per neuron, the size of the disc would be unimportant in assessing RNFL thickness. However, eyes with larger optic discs are known to have more axons in histological studies of monkeys,19
and disc rim area increases with increasing disc area in humans,20
also indicating that there may be more axons in larger discs. Our findings are consistent with this hypothesis. The current study also found that larger eyes and more myopic eyes had a statistically lesser mean RNFL thickness in a model that adjusted for disc area. Thus, highly myopic or long eyes inherently may have fewer ganglion cell axons than emmetropic average-size eyes. Alternatively, eye length and refraction could be related to another variable, such as disc area. In this case, to be consistent with other findings, greater axial length would have to be associated with smaller disc area, but in our model, there was no interactive relationship between axial length and disc area. Another possibility is that OCT measurements are affected optically by greater axial length and higher myopia, producing apparently thinner RNFL as an artifact. A longer eye will produce a larger scanning circle diameter, thereby measuring the RNFL in an area thinner than that intended. Whatever the reason for these differences, axial length or refractive error may need to be taken into account in the assessment of RNFL thickness by OCT.
We found a significant relationship between optic disc area and RNFL thickness. Eyes with larger optic disc areas had thicker peripapillary RNFL measurements, independent of axial length, refractive error, and other demographic and clinical variables. Savini et al found a similar relationship between optic disc size and peripapillary RNFL thickness in 54 normal Caucasian subjects using a model that took into consideration age and several optic disc parameters.21
There are several possible explanations for this finding. First, it is likely that eyes with larger optic discs have more axons. A histologic study in monkey eyes demonstrated this,15
although a human histologic study failed to find a relationship between optic disc size and RNFL thickness.22
A second hypothesis is that eyes have the same number of axons regardless of disc size but that the fixed circular scan of the peripapillary RNFL measurements performed with Stratus OCT somehow produces a thicker measurement. Indeed, in larger discs the OCT measurement is made closer to the optic disc margin, where axons may be sampled as they are at a different incident plane to the scanning beam. If they are coursing more obliquely (either heaping up into the disc rim or already diving downward into it), the measurement by the instrument might give a thicker value than in eyes in which the axons are further from the disc rim and are completely perpendicular to the measuring light. A histologic study in normal human eyes showed that RNFL thickness decreases with increasing distance from the disc margin.18
Regardless of the explanation for this finding, it suggests that somewhat greater predictive power in the assessment of RNFL might be achieved if disc area were taken into account, perhaps by measuring the optic disc area using the Stratus OCT optic disc scan protocol and incorporating this variable into the comparison with the normative database. Another way to adjust for this would be to measure RNFL at a fixed relative distance from the optic disc margin. Carpineto et al23
performed Stratus OCT measurements using the fixed 3.46-mm-diameter circle and a variable scan circle diameter designed to measure the RNFL 0.85 mm from the edge of the optic disc in the same group of 30 normal eyes. They confirmed the findings of the current study—namely, that eyes with larger optic nerves had thicker RNFL measurements—but also showed that RNFL thickness is unrelated to vertical optic disc diameter if one measures RNFL thickness a fixed distance from the disc margin.
Other demographic and clinical factors, such as right or left eye and gender, were found not to be important determinants of RNFL thickness. Thus, these factors need not be considered when determining what constitutes normal RNFL measurements. Other factors not explored in the current study, such as time of day and IOP, may need to be evaluated further for their effects on RNFL thickness.
The purpose of this article was to identify factors that may merit consideration in future normative databases, and not to present normative data for general clinical use with Stratus OCT or any other product. The analyses presented here were done independently of any work performed by the manufacturer in calculating the normative data present commercially in its Stratus OCT product. Statistical estimation methods differed, and it is therefore likely that the values presented in this article differ somewhat from those in the commercial Stratus OCT software.
It is important to note that the first and fifth percentiles of normality described in the Fast RNFL printout, which are based on the data collected in the current study, are not necessarily diagnostic of glaucoma. First, 1% and 5% of normal individuals would be expected to have values for RNFL thickness in the bottom first and fifth percentiles, by definition. Second, the current version of the Stratus OCT software adjusts for age but not ethnicity, axial length, or optic disc size. Based upon the present study, adjustment for such parameters in current and future iterations of OCT technology would be expected to provide better sensitivity and specificity for glaucoma detection. Additional normative data may need to be collected that include more subjects of African, Asian, and Indian–Asian ethnicity and those with higher degrees of refractive error.