advertisement

Editors Selection IGR 9-4

Intraocular pressure: Night-time IOP

Comment by Christopher Bowd on:

19694 Use of a Bayesian network to predict the nighttime intraocular pressure peak from daytime measurements, Nordmann J-P; Berdeaux G, Clinical Therapeutics, 2007; 29: 1751-1760

Find related abstracts

In a significant proportion of glaucoma patients, visual sensitivity continues to deteriorate despite an acceptable IOP measured during clinic visits. It is possible that these patients exhibit nighttime peaks or fluctuations in IOP not detected by tonometry during office hours and that these undetected changes contribute to glaucomatous progression. The paper by Nordmann and Berdeaux (1253) attempts to use a Bayesian classifier to predict nighttime IOP measurements from daytime measurements. Daytime IOP was measured at 0800 hrs, 1200 hrs, 1600 hrs, and 2000 hrs and night-time IOP was measured at 2400 hrs and 0400 hrs (although various eyes were missing various measurements, partly because data were pooled from three somewhat similarly conducted clinical trials, yielding a total of 103 participants). At each time point, IOP was dichotomized as 'controlled' (arbitrarily, ≤ 18 mmHg) or 'uncontrolled' (> 18 mmHg) and, using commercially available Bayesian analysis software, the chance of nighttime IOP control was predicted based on control or un-control at each daytime measurement.

Results indicated that a patient with controlled IOP at both 1200 hrs and 2000 hrs had the best chance (79.8%) of controlled IOP during nighttime measurements. If IOP was uncontrolled at 1200 hrs or 2000 hrs, the chance of nighttime control decreased to 30.0% and 18.7%, respectively. If IOP was uncontrolled at 2000 hrs, the chance of nighttime control was only 5%. Unfortunately, because the authors pooled data from several studies involving different drugs and different treatment regimens, it is unclear to which scenario the conclusions are most applicable. In addition, much of the Results section of the manuscript is devoted to describing differences in results between groups and this, presumably, was not the main point of the study. Finally, arbitrarily dichotomizing the measurements resulted in the loss of much of the information gathered. Results from this study can be improved with a well-controlled, prospective trial.

• ← Previous
• Next →