When the U.S. Food and Drug Administration in 2010 issued a draft guidance on developing drug combinations, it was written primarily with anti-infectives and cancer drugs in mind. Alzheimer's disease clinicians who have read the guidance consider it too vague to design trials with it. At an ACT-AD/C-Path meeting held in November 2012 (see Part 1 of this series), Bob Temple and Rusty Katz of the agency explained it further in a discussion with the audience.
Since the 1970s, companies have had to show that each drug makes a contribution to the claimed effect of a given combination. Dozens of blood pressure-lowering and antibiotic combinations were developed that way. “You have to have reason to believe each component is contributing; otherwise, you are only getting the risk of side effects for no benefit,” Temple said. In recent years, however, scientific progress on underlying disease pathways has loosened this requirement somewhat to where strong biomarkers and pathophysiology knowledge can make the case that each drug contributes. This works best when molecular disease mechanisms are well understood.
For combinations of unapproved drugs, researchers generally have much less safety and dose response information on the single drugs than they do for approved drugs. This complicates testing combinations but should not deter AD researchers, Temple said. If the drugs have added effects in a disease as bad as Alzheimer’s, then making combination development as efficient as possible is truly urgent. “It’s of potentially enormous value,” Temple said.
Even with little information on the drugs, developing a combination makes sense if there is a compelling biological rationale, for example, if each drug hits a distinct target in the same molecular pathway. If a drug cannot win approval by itself because it only shows an effect in combination, then that is a valid reason as well, Temple said. In breast cancer, trastuzumab, better known as herceptin, falls into this category. By itself it is ineffective, but as a combination drug, it changed breast cancer therapy. Herceptin might not have been found if it were only studied individually. In AD, scientists believe some drugs cannot be developed singly because trials take too long, especially in early stages when the disease worsens slowly and drug effects are subtle. In the absence of strong individual drug effects, a preclinical model could show an additive or synergistic effect of a combination, and a short in-vivo study on biomarkers can establish that the combination is active and has additive benefits, Temple said.
The draft guidance mentions synergy, that is, a greater-than-additive benefit from the combination. That point can be taken with a grain of salt, Temple said. In reality, few established combination therapies in other indications are synergistic. Most are additive. “It is very hard to show synergy. Additive is good enough,” Temple said.
Toxicity and pharmacokinetics would have to be done in the usual ways, plus some research on whether there are drug-drug interactions.
Traditionally, combination trials have shown the contribution of single versus combination therapy with a factorial study. This type of trial design compares the combination with each of the component drugs, all added to standard of care or compared to placebo. One challenge in AD is that blinding such a trial may be tricky if one drug comes in a pill and the other requires infusion.
Factorial designs with adaptive features would be welcome, Temple said. That is, in part, because adaptive trials are well suited to testing several different doses in addition to combinations. Factorial designs can get unwieldy, requiring many arms if the trial starts out with different drugs at different doses. Helping matters, though, is that adaptive trials can drop those arms that underperform on the interim endpoint. This endpoint can be pharmacodynamic. In cancer, some factorial combination trials start out with more than 10 arms, but aggressive interim analysis allows scientists to trim that number within a few weeks or months into the trial.
How well that will work in AD remains to be seen. “The short answer is, if the effect is dramatic, all this works fine. If the effect is small, it’s going to be hard,” Temple said. The great hope is that preclinical and human Phase 1 and Phase 2 do show a contribution by each component. That is because in this situation, Phase 3 could compare only the combination to placebo or standard of care, obviating the need for a factorial Phase 3 that would have to be multiple times as large. “If the effect size is clearly larger for the combination than either of the components, then you have a much smaller Phase 3 study and enormously increased power,” Temple said.
In essence, the news at this point about FDA’s view on combination trials in AD is that the agency intends to be flexible in how sponsors can demonstrate evidence of individual drug contribution. “If a combination does something wonderful, what are we going to do—not approve it because it did not follow all the rules?” Temple said. As an example, he reminisced about his early days as a physician intern. “We’d give people with acute leukemias a drug and [the leukemias] would go away; three months later it was back and they died. Then people developed combinations, and all of a sudden it did not come back. Did we know which drug contributed exactly what? No. You did not need to. You knew the combination made them better.”
A major effect in Alzheimer’s would overcome significant safety problems as well, Temple added.
That said, Katz urged trialists to consult the FDA before going ahead with combination trials. “Everything depends on the details of the drug and the scenario. You have heard me say that before and it’s especially true here. We are very interested in developing combinations,” Katz said, “But there is no precedent. There is no standard case that you could apply the guidance to and be done. You really have to talk to the division.”
Over the past two years, adaptive trials and Bayesian statistics have become a staple of discussions on how to make Alzheimer’s clinical trials cheaper and more successful (see ARF related news story). At the ACT-AD/C-Path meeting, both Temple and Katz encouraged the field to consider such designs and, indeed, some trials are already applying them (see ARF CTAD story). At the ACT-AD/C-Path conference, researchers brainstormed on how they could launch clinical trials testing several unapproved experimental therapies at once. Adaptive trials guru Don Berry of the MD Anderson Cancer Center in Houston, Texas, argued that combination trials in AD should incorporate adaptive features, and that adaptive designs can be used by several otherwise competing companies in a shared infrastructure (see Part 3 of this series).
Berry reiterated that more than 300 adaptive trials in cancer at his institution have shown that looking at the data during the trial and adjusting the trial accordingly provides better information with fewer patients and allows a single trial to answer more than one question. On combination treatments, the tendency of tumors to become resistant to a single drug drove innovation toward combination trials as early as the 1980s. A 1993 breast cancer trial answered the question of whether increasing an effective dose of doxorubicin would add benefit (it did not) and whether taxol added benefit over doxorubicin alone (it did). This historic trial used a 3x2 factorial design of three doses of doxorubicin against one dose of taxol or no taxol (Henderson et al., 2003).
In Alzheimer’s trials, endpoints are such a wide open question at this stage that a combination trial could be engineered to address it head on, Berry said. A trial could determine whether drug A affects a different endpoint than does drug B or the combination. It could test whether effects on different short-term biomarkers create synergism toward a benefit on a later clinical endpoint. And it could test if particular patients respond to drug A and patients with perhaps a different biomarker or genetic profile respond to drug B or the combination.
Neil Buckholtz of the National Institute on Aging in Bethesda, Maryland, said that, to his mind, the critical point for succeeding with adaptive combination trials is to first find biomarkers that change, with sufficient measurement precision, in a practicable time frame. He urged that before large factorial studies are undertaken, small studies be done that dispense single drug and combinations over short periods of time and quantify which drug moves which biomarker, and how soon after dosing.—Gabrielle Strobel.
- Henderson IC, Berry DA, Demetri GD, Cirrincione CT, Goldstein LJ, Martino S, Ingle JN, Cooper MR, Hayes DF, Tkaczuk KH, Fleming G, Holland JF, Duggan DB, Carpenter JT, Frei E, Schilsky RL, Wood WC, Muss HB, Norton L. Improved outcomes from adding sequential Paclitaxel but not from escalating Doxorubicin dose in an adjuvant chemotherapy regimen for patients with node-positive primary breast cancer. J Clin Oncol. 2003 Mar 15;21(6):976-83. PubMed.