We read with interest the paper published by Payne et al.,^{ 1 } in the November 2011 issue, in particular the statement that the ratio of OCT thicknesses measured at different time points is machine invariant because of the cancellation of the baseline thickness. We believe that this statement is incorrect, because of the nature of the calculation of retinal thickness using OCT. We present the reasons below.

The retinal thickness is the distance between the inner retinal surface and the reference plane. This distance represents interval data, as there is no natural zero, and the retinal thickness assumes a numerical value only when the reference plane is defined. As a result, the retinal thickness value contains information about the reference plane for each particular machine.

Consider two OCT machines A and B, each of which measures from different reference planes,

*a* and

*b*, respectively. Let plane

*c* be the position of the inner retinal plane and the retinal thickness

*d* be measured at two different times,

*t* _{1} and

*t* _{2}. During this time interval, let the position of the retinal layers change because of the effect of disease or intervention (

Figs. 1A,

1B).

If the ratio of two measurements performed at different times is independent of the OCT machine used, then these values will be equal in the same individual.

In the general case, at time

*t* _{2} it is best to assume that the positions of all retinal layers have changed, so that

Thus, the ratio of thicknesses measured at times

*t*_{1} and

*t*_{2} for both machines is

and

The condition for machine invariance (i.e., the ratio of two thickness measurements made at different times on different machines) is

However, it is clear that the values of retinal thickness derived by the two machines will be different, because the planes *a* and *b* are different and also because the planes *a* and *b* may have moved from their original positions at time *t* _{2}.

The difference between the ratios can be 0 only if the condition *a* = *b* is applied, but this is not the case with the various OCT machines.

Therefore, neither the ratio of the thicknesses nor the log ratio can be machine invariant. This observation is illustrated by the example below. Assume that the reference planes are 50 μm apart, so that

*b* −

*a* = 50 μm, and they do not move during the time interval:

As an alternative to ratio calculation, we can also explore whether the linear difference between thickness measurements is machine invariant. Consider the difference between the thicknesses at the different time points:

and, similarly,

The condition for machine invariance here is that

From this calculation, we have

which is the case only if the reference planes do not change position during the time interval or if the difference between the two planes is negligibly small.

It may be that for some diseases, such as diabetic retinopathy and other retinal vascular diseases, that only the inner retinal layer moves and the reference planes (e.g., the tips of the outer segment or the first bright RPE reflection) do not move, in which case the linear difference between measurements will be machine invariant [as (*b* _{2} − *b* _{1}) is zero and (*a* _{2} − *a* _{1}) is zero].

However, in florid cystoid macular edema, where the photoreceptor layer itself detaches from the RPE or where the position of the RPE layer moves (e.g., RPE detachment reducing after intravitreal ranibizumab), then the linear difference between different machines will not be invariant.

In summary, retinal thicknesses derived from OCT measurements are interval data. There is no natural zero, and the numerical value of the thickness always depends on the choice of reference plane. The ratio of two thickness measurements made at different time intervals and the logarithm of this ratio varies from machine to machine. The linear difference between thicknesses at these different times may be machine independent, but only if the position of the reference plane remains unchanged with the passage of time.

We therefore warn the retinal community against performing studies in which different OCT machines are used and ratios of thicknesses are calculated and treated as comparable. Such comparisons would introduce errors in the data that are collected. We hope that our analysis explains this concern.