advertisement

---

Long-Term IOP Fluctuation May not be as Important as Previously Reported

Kuldev Singh

Editorial

Comment on the editorial by Paul Palm-berg in Ophthalmology (2007; 114: 203-204), who discussed the paper by Bengtsson et al. on 'Fluctuation of intra-ocular pressure and glaucoma progression in the early manifest glaucoma trial' in Ophthalmology (110).

Over the past decade, multitudes of post hoc analyses have been derived from large prospective glaucoma studies. Many of these analyses come from trials that are not well suited to answer the 'after the fact' questions posed by those who have access to data. The reports from such analyses have the potential to mislead practitioners about relationships between IOP and glaucomatous optic nerve damage.

Bengtsson et al. are thus to be congratulated for performing a post hoc analysis that actually makes sense. They have elegantly shown that an evaluation of all patients in EMGT reveals that the mean IOP is a strong risk factor for glaucoma progression and that inter-visit fluctuation is not. The authors have highlighted the fact that the propensity for IOP fluctuation increases with higher mean IOP and that this relationship needs to be addressed and corrected for when conducting a risk factor analysis of fluctuation.

The authors also correctly point out that subjects in whom treatment is advanced when progression occurs are no longer appropriate for assessing the relationship between fluctuation and progression. Progression can lead to treatment changes which induce IOP fluctuation, thus making it difficult to establish causation. Did the fluctuation cause the glaucoma to progress or did progression lead to further treatment which induced inter-visit fluctuation? EMGT is particularly well suited for the analysis performed by the authors as treatment was generally not advanced based upon an IOP threshold during the course of the study.

The editorial by Palmberg which accompanies this paper nicely highlights the views of the author with regard to the concept of target IOP, but does not fully acknowledge the quality and significance of Bengtsson et al.'s work. The data requested from the EMGT investigators towards the latter part of the editorial is unrelated to the topics addressed by the authors and once again, relates primarily to the target IOP concept, the relevance of which remains controversial.

One common misperception within the medical profession is the belief that large well designed multi-center clinical trials sponsored by independent funding agencies are not subject to bias. Such studies create large data sets which, particularly in the case of post hoc analyses where 'new questions' are created during the study, can be sliced and diced in an effort to prove prevailing preconceived notions regarding how a disease 'should' behave.

Good science begins with a hypothesis which is then tested in an experiment. The results of the experiment lead to conclusions which may or may not support the original hypothesis. With large data sets generated from clinical trials, the potential exists for the conclusions to precede the analysis. Since there are infinite different ways to look at such data sets, one can come up with virtually any conclusion and then try to find the data from the study to support it. For these reasons, it is best to focus on the initial study hypothesis which, in most instances, is the one that the study is best designed to address.

While it is generally best to avoid post hoc analyses, the diligence and integrity displayed in the work by Bengtsson and colleagues in this paper, as well as a previous similar report from the Malmö Ocular Hypertension Study, should be applauded.