[The following letter was published in the peer-reviewed journal "Clinical and Experimental Ophthalmology" in July 2008. The citation is:
Leffler CT, Hennessy A, Farukhi Y. Current state of the one-eye trial of glaucoma medications. Clin Experiment Ophthalmol 2008;36:492-3.]
Current state of the one-eye trial of glaucoma medications.
Christopher T Leffler MD MPH, Amy Hennessy MD MPH and Yousaf Z Farukhi BS
Department of Ophthalmology, Medical College of Virginia Campus, Virginia
Commonwealth University, Richmond, VA, USA
Copyright Journal © 2008 Royal Australian and New Zealand College of Ophthalmologists.
--------------------------------------------------------------------------------
Received 14 May 2008; accepted 22 May 2008
Dayanir et al. studied the one-eye trial.[1] From the standpoint of effectiveness, the recent focus,[1–5] only one reason for uniocular trials has been offered by Drance or others – the reduction of spontaneous fluctuation unrelated to treatment by using the contralateral eye (A) as a control.[4,5] The initially treated eye (B) response between the baseline (1) and trial visit (2) is classically estimated:
(B2 − B1) − (A2 − A1).
Dayanir et al. and others[2,3,5] have tried to assess whether the response in each eye is similar. This focus relates to the (erroneous) assumption[3] that uniocular trials require fellow eye response similarity. This assumption was not stated explicitly by Drance or Smith. Even with dissimilar fellow eye responses, the first-treated eye can be used as a control to assess the second eye IOP change between the uniocular trial visit (2) and a binocular use follow-up visit (3):
(A3 − A2) − (B3 − B2).[2] Alternatively, on visit 3 the drug can be used in eye A only, and the eye A effect is: (A3 − A1) − (B3 − B1).
Using the contralateral eye as a control in the one eye trial assumes: (i) minimal cross-over effect to the contralateral eye, and (ii) a high correlation of spontaneous fluctuation between the two eyes. However, any drug has a large apparent cross-over effect if the patient takes the drug in both eyes or the wrong eye. Spontaneous IOP fluctuation in each eye has components which are correlated and uncorrelated. The contralateral eye control eliminates correlated fluctuation, but exaggerates uncorrelated fluctuation. The variance of the difference of two random variables is the sum of the variances of each variable. A difference of differences involves four times the uncorrelated variation of any individual IOP reading.
Whether cross-over effects and uncorrelated variation are large enough to render invalid the subtraction of the IOP change in the contralateral eye is an empiric question. The broader question is how the uniocular trial IOP measures (A1, B1, A2,
B2) predict (i) the likely future treated IOP values (A3 and B3), and (ii) treatment response. For treated IOP, using the mean IOP over several visits improves precision.[4] The reference standard for medication effectiveness is the IOP difference over multiple visits at which the patient has been randomized to treatment or placebo. This n-of-1 trial4 has never been conducted for glaucoma uniocular trials. The best currently available effectiveness measure is the mean follow up minus the mean baseline IOP.
No evidence supports subtraction of the contralateral IOP change. One multivariable analysis found that the regression coefficient for the independent variable
(A2 − A1) for prediction of A3 and B3 was generally close to zero or positive (not negative).[4] Subtraction of the contralateral IOP change slightly decreased the correlation of uniocular trial response in fellow eyes from 0.102 to 0.097.[5] The higher correlation of (A3 − A2) − (B3 − B2) with (B2 − B1) − (A2 − A1), (as compared with (A3 − A2) versus (B2 − B1)), was thought to support classic teachings.[2] However, the former equation has (B2 − A2) on both sides. Even random numbers would correlate. Only one study compared uniocular and binocular trials in the same patients. Both trials had equivalent information content.[4]
The main reason to perform a uniocular trial is safety. Half the systemic dose is achieved and only one eye suffers local side-effects, which are more easily diagnosed.
Either uniocular or binocular trials may be performed. In uniocular trials, no evidence supports subtracting the untreated eye IOP change. Multivariable regression on existing datasets[1,2,5] can confirm how uniocular trial IOP values (A1, B1, A2, B2) predict follow-up IOP, and whether uniocular and binocular trials have similar information content.[4]
Christopher T Leffler MD MPH, Amy Hennessy MD MPH and Yousaf Z Farukhi BS.
References
1. Dayanir V, Cakmak H, Berkit I. The one-eye trial and fellow eye response to prostaglandin analogues. Clin Experimental Ophthalmol 2008; 36: 136–41.
2. Chaudhary O, Adelman RA, Shields MB. Does the response to initial glaucoma therapy in one eye predict the response in the fellow eye? Invest Ophthalmol Vis Sci 2007; 48: E-Abstract 5556. [ARVO abstract #B895].
3. Realini T, Vickers WR. Symmetry of fellow-eye intraocular pressure responses to topical glaucoma medications. Ophthalmology 2005; 112: 599–602.
4. Leffler CT, Amini L. Interpretation of uniocular and binocular trials of glaucoma medications: an observational case series. BMC Ophthalmol 2007; 7: 17.
5. Takahashi M, Higashide T, Sakurai M, Sugiyama K. Discrepancy of the intraocular pressure response between fellow eyes in one-eye trials versus bilateral treatment: verification with normal subjects. J Glaucoma 2008; 17: 169–74.
