The Tube versus Trabeculectomy (TVT) Study was a multicenter randomized clinical trial that compared the safety and efficacy of tube shunt surgery and trabeculectomy with mitomycin C in glaucomatous eyes that had previously undergone cataract surgery or glaucoma surgery.
Over the past fifteen years, the Tube Versus Trabeculectomy Study Group has published 14 papers describing the design, execution and findings of their study. Similar IOP reductions and use of medical therapy were observed with both procedures after five years of follow-up. Early postoperative complications occurred more frequently after trabeculectomy, but both procedures had similar rates of late postoperative complications and serious complications at five years. No significant difference in visual outcomes was seen.
In this most recent paper, the TVT Study Group evaluated Visual Field (VF) findings in 122 eyes of 122 patients, with 61 eyes in each treatment group. A total of 724 Humphrey 24-2 VFs were taken during the study. However, 140 VFs were excluded because they did not satisfy the study's criteria for FP and FN rates. An additional 148 VFs were excluded on the basis of central visual acuity levels or changes in VA compared to baseline, leaving 436 VFs covering more than three years of follow-up in the two treatment groups together - and an average of 3.6 visual fields per eye. Both treatment groups had average baseline MDs of approximately -13 dB.
The authors found MD rates of change of -0.60 dB/year in the tube group and - .38 dB/year in the trabeculectomy group, which were not significantly different (P = 0.34). This naturally raises two questions: What progression rate difference was this study powered to detect, and what difference might be deemed clinically important? The authors point out that VFs were a secondary outcome measure, and thus power calculations were not prospectively completed. However, as a post hoc analysis, the authors determined that, '[…] if there were to be a significant difference, we can assert with 95% confidence that the difference is no more than 0.2 dB/year faster in the trabeculectomy group and, similarly, no more than 0.7 dB/year faster in the tube group.' I asked the TVT Study team for clarification, and here is their reply: 'While a 95% confidence interval suggests the per year difference between treatment groups in MD progression rates is < 1 dB, the potential differences become clinically significant when extrapolated out to longer follow up. For example, after five years, the MD progression rate could be 3.5dB faster in the tube group or, alternatively, 1dB faster in the trabeculectomy group.'1
In this study, cut-off values for perimetric reliability parameters were based on a recent paper by Yohannan et al.2 On the basis of that paper, the TVT authors quite reasonably chose to reject tests having false positive rates > 20% or false-negative rates > 35%, and to not consider FL rates. However, use of these limits led to rejection of 19.3% of available visual fields. Three papers, when considered together, suggest to me that reliability parameters may be the least reliable metrics produced by the Humphrey perimeter,3,4,5 leading to the conclusion that perimetric test results should seldom be discarded solely on the basis of reliability parameters.6 Given that this paper's most important finding may regard the difference between tube and trabeculectomy rates of perimetric progression and not the absolute rates found in each group, and noting that patients really have been randomized in this study, should we not consider how treatment group rates of progression compare when reliability limits are not imposed? Similar discussions might also be productive, regarding the 20.4% of fields that were discarded due to visual acuity requirements.
We congratulate the TVT Study Group on fifteen years of key contributions to glaucoma surgical care and look forward to further discussions on this important aspect of their work.
Reliability parameters may be the least reliable metrics produced by the Humphrey perimeter,3,4,5 leading to the conclusion that perimetric test results should seldom be discarded solely on the basis of reliability parameters