What Is the Value of Clinical Trial Subgroup Analysis?
Opinions vary on the extent we should use these data to guide clinical decision-making.

Dr. Fine: We've talked about how we don't have as much data as we'd like to have from the pegaptanib sodium (Macugen) clinical trials regarding large, minimally classic or occult lesions. A biostatistician will tell you the first and most important comparison in any clinical trial is always all the patients assigned to one group vs. all the patients assigned to the other group. However, subgroup analyses can shed further light on those comparisons.

For example, in the initial Treatment of Age-related Macular Degeneration with Photodynamic Therapy (TAP) study report¹, the classic-only lesions group carried the study. A table in the initial TAP publication in the Archives of Ophthalmology shows when any occult choroidal neovascularization (CNV) was present, there was no treatment benefit. A different committee might have recommended verteporfin exclusively for classic-only CNV.

Dr. Williams: The TAP study also showed an overall treatment effect. If you consider the primary endpoint most important, you could make the argument that although some subgroups drove the results, the classic-only CNV subgroup was not predefined. The predefined subgroup was predominantly classic CNV, not classic-only CNV. That's really a subgroup of a subgroup.

Dr. Fine: But there were substantial differences between classic only and the other subgroups.

Dr. Williams: The differences were enormous, but subgroup analyses of subgroups are fraught with danger. You can dissect any study as finely as you want. Although it appears classic neovascularization drove the results to some extent, I'm not sure I'd accept that only purely classic lesions benefit from PDT.

Dr. Fine: It doesn't necessarily alter how I use PDT, but it alters my expectations and how I talk to patients about expected treatment outcomes.

A parallel example, from quite some time ago, is the use of photocoagulation for juxtafoveal neovascular lesions in AMD patients with hypertension.

The overall results of the krypton photocoagulation study² showed a treatment benefit, but the subgroup with hypertension had no treatment benefit.

You could use these data to decide you won't treat this hypertensive group. Or, because these patients didn't do worse and some did better, you could treat a patient who has hypertension as long as you explain your decision so expectations as to possible outcomes are appropriate.

Dr. Slakter: Some people argue that we should never use verteporfin to treat anything but purely classic lesions because the TAP study showed it worked only in classic lesions. I think we all agree this is an interesting finding, but when a study is powered to detect a difference in a group and then you take a subgroup out and say it didn't show a statistical difference, that doesn't mean there wasn't one. To say we should use verteporfin only in classic-only cases would be twisting the science and the data.

Dr. Klein: In comparing treatments and making treatment decisions, I agree we should consider overall treatment effect rather than results of subgroup analyses. Comparing study results is difficult for several reasons, including those already mentioned, and comparing results of subgroup analyses even more so.

For example, I don't know that we can necessarily assume pegaptanib is superior to PDT for large, minimally classic or occult lesions, based on the available subgroup information. When visual acuity differences between verteporfin-treated and placebo groups are plotted against lesion size, we see a treatment effect for predominantly classic lesions of all sizes but only for smaller occult and minimally classic lesions.

One contributing factor may be that while visual acuity decline with lesion size is similar for the three verteporfin-treated CNV subtypes, in the placebo or natural history groups, visual acuity is much more significant than with the other two subtypes. We don't know whether pegaptanib would improve on this relatively less severe vision loss in larger minimally classic or occult lesions. Even if we had the results for this subgroup, it is quite possible that numbers may be inadequate for us to reach any conclusions.

Dr. Slakter: One advantage we have from the verteporfin data set is that we actually have information showing we should not use PDT in patients who have large occult lesions and good vision. We have the data that tell us not that there may or may not be a treatment benefit, but that there appears to be a treatment negative.

From a treating physician's point of view, that's important. I don't need statistically significant treatment positives for every subgroup, but I'd like to know if I'm likely to do harm. We don't have that with the available pegaptanib data.

Dr. Reichel: The pegaptanib data we've seen indicated that 20% of patients in the small-lesion group with good visual acuity and, presumably, early lesions showed three lines of improvement in visual acuity vs. zero in the sham group.

For the overall group, I believe the numbers are 6% showing three lines of improvement vs. 2% in the sham group. That seems to be driving the improvers, and that subgroup may be the one that achieves the most benefit from pegaptanib.

Dr. Slakter: Subgroup data is interesting and should give us information, but we should not use it to justify particular treatment decisions. That's what we should take away from this discussion.

REFERENCES

1. Treatment of Age-Related Macular Degeneration with Photodynamic Therapy (TAP) study group. TAP report No. 1. Arch Ophthalmol. 1999;117:1329-1345.

2. Macular Photocoagulation Study Group: Krypton laser photocoagulation for neovascular lesions of age-related macular degeneration. Results of a randomized clinical trial. Arch Ophthalmol. 1990;108:816-824.

 

Dr. Johnson: Patients could have at least three potential responses to pegaptanib. In some patients, all of the fluid might disappear and vision might improve. In these cases, I'm likely to inject every 6 weeks. After about three treatments, I may be tempted to withhold an injection, follow closely and resume treatment if leaking resumes.

Some patients may not respond at all. For me, it would be difficult to perform more than two or three treatments before concluding that repeated injections aren't beneficial. In the middle, we will likely have patients with stable vision who show low-grade leakage evidenced by optical coherence tomography (OCT) and clinical examination. In these cases, we'll have difficulty deciding on our next step. As long as vision is stable, the tendency will be to continue treatment on the assumption the patient might be doing worse without treatment, but I don't have a good sense of how long to persist.

Dr. Klein: That's exactly what I would do. I also think it's important to follow these patients with OCT. I plan to image with OCT every 6 weeks along with the injection and do fluorescein angiography every other visit.

Dr. Chow: My big fear with the current treatment protocol is the results I'll get if I don't inject every 6 weeks. In fairness, we could say the same thing about verteporfin. In clinical reality, most retina specialists deviate from the TAP protocol, treating far less than in the trial. The question is what visual results are we getting when we deviate from the protocol? This and other questions are being answered with the information coming from the InSight Registry, which can be found on the Visudyne Web site (www.visudyne.com). The Registry is an interactive online database that enables us to record individual verteporfin patient data while offering the opportunity to view and compare collective clinical data.

Dr. Williams: Even people who advocate following the pegaptanib protocol exactly are somewhat selective. If you're going to be dogmatic about the protocol, don't talk about nonsubfoveal lesions. Don't talk about occult lesions that are not active or lesions larger than 12 disc areas. If you want to follow the protocol, you can apply it only to the patients who were studied. This takes me back to my point that none of these clinical trials was handed to us on stone. They provide very useful information, but if we adhere to precisely prescribed treatment regimens, we'll be throwing away physician judgment, which would be a mistake.

Dr. Fine: We continue to follow some protocols without variation for many years after publication of clinical trial data. Diabetic retinopathy is one example. One reason is that the treatments are effective.

In the case of verteporfin, we've changed our approach because we have a technology we didn't have before: OCT. So we've moved away from the recommendation of re-treating when we're not sure if there is leakage. If we don't see fluid on OCT, it almost doesn't matter what the fluorescein angiogram looks like. We aren't going to treat.

We have new information, and that's a convincing, justifiable reason for departing from a protocol. In the absence of new technology that provides new information, or compelling evidence to not follow a protocol, I think we're duty-bound to follow the protocol.

Dr. Klein: No matter what the data show, we will follow these patients to try to figure out what we should be doing

PRACTICAL AND PRACTICE CONSIDERATIONS

Dr. Brucker: Let's say you have a big practice. Do you want to bring your patients in every 6 weeks, give them a pegaptanib injection and not think about what you're doing for 2 years? Or do you want to bring patients into the office, work them up, dilate them, evaluate them with a fluorescein angiogram and perhaps an OCT, and then make your decision whether or not to re-treat them with a pegaptanib injection?

Dr. Klein: Emphasis should be on the cost and convenience for the patient, if the treatments are roughly equal in terms of efficacy and safety. I'm not dismissing the problems we'll have in our offices, but somehow we'll adapt.

Dr. Slakter: We're going to be dealing with an office volume phenomenon. When the number of patients coming in every 6 weeks starts to increase by a factor of the number of patients treated every week, we'll see a huge volume increase. If we had a treatment given every 6 weeks that improved vision by three lines in 80% of patients, we wouldn't be debating whether we had to bring patients back every 6 weeks. We'd likely bring them back, check their vision, and if it were stable and getting better, we'd give them their injection and assess the situation in a year.

But that isn't what we're dealing with. We're dealing with a treatment that's been shown to be better than placebo. If seven of 10 patients treated with pegaptanib are doing statistically better than nothing, does that mean three of 10 are not responding to therapy?

We have to figure out how to identify these patients earlier and when to change their therapies. We don't have answers at this point.

Dr. Williams: We have to remember how patients are viewing this. They're receiving intravitreal injections or PDT and many are still losing vision. The cost-benefit ratio will drive the time frame of patients' therapy.

They don't care about three lines this way or that way; they just know how treatment is affecting their lives. Many are going to say, "I'm sure you're right, doctor; this is a great treatment, but it's not working for me, and it's very disruptive to my family to come here so many times a year."

That will be a huge issue. We may have patients dropping out of certain therapies as often as their doctors do.