This spring, the Duchenne muscular dystrophy community rejoiced when PTC Therapeutics of South Plainfield, N.J., announced that an experimental drug aimed at the disease had passed safety tests in human volunteers and would now be tested in boys with DMD.

This drug, like many others, was successfully tested in mice lacking dystrophin, the protein that’s also missing in the human form of the disease.

But scientists studying DMD are also aware that other treatments that have looked good in the dystrophin-deficient mouse haven’t done as well when tried in people.

So, what can mice — let alone fruit flies, worms, fish and a few cells — tell us about human disease?

Can a mouse with a genetic mutation that causes the same protein deficiency that exists in Duchenne MD, but not the same muscle weakness, predict how children with the disease will respond to an experimental drug?

As with most complex problems, the answers are yes, no, maybe and sometimes.

Can't You Just Test It in People?

Sometimes people with rapidly progressive, life-threatening conditions want to know why they can’t volunteer to be "guinea pigs" for compounds that haven’t yet been tested in any organism. They feel they have "nothing to lose."

To that question, Robert Temple, director of the Office of Medical Policy at the U.S. Food and Drug Administration, answers that there are treatments that can make a disease worse, as well as those that might make it better or not change it at all.

"We believe that even someone with a dreadful disease probably doesn’t want to die tomorrow," Temple says.

	Fruit flies, such as this one, as well as worms, fish and squid, can answer some questions about drug effects in humans.

Although there are situations in which drugs are fast-tracked into human testing, establishing some minimal safety assurance in animals is always required for a new drug in the United States. That doesn’t mean animals are perfect replicas of humans and their diseases, or that harm resulting from a new drug can be completely predicted and avoided. It’s just a minimum standard to which U.S. researchers are expected to adhere.

As they say about democracy, "It’s not perfect, but it’s the best form of government we have." That’s the way most researchers and government regulators feel about animal and other lab-based "models" of human disease.

"The amount of animal data initially required before studies can be conducted in human beings is quite modest," Temple says, adding that the extent to which animal research slows drug development is small.

Safety First

Typically, when a company applies for approval for a new, previously untested drug, the FDA requires short-term testing in animals prior to beginning human testing, just to make sure there are no glaring safety issues.

Long-term testing in animals then generally continues and may overlap with the first human test, called a phase 1 clinical trial. Phase 1 trials are small and are designed to measure safety and toxic effects, sometimes in healthy volunteers and sometimes in people with the targeted disorder. (Test compounds that have a high likelihood of injuring healthy volunteers but may help patients — for instance, those that change genetic information — are studied in affected patients, even in phase 1.)

Phase 2 studies continue to test safety but also look at the effectiveness of the compound in people with the disorder. Phase 3 studies are large-scale trials in people.

Russell Katz, director of the FDA’s Division of Neuropharmacological Drug Products, admits that animal and cellular models aren’t perfect replicas of diseases. But, he says, "There has to be some minimum attention paid to figuring out what the risks might be before we study the human experience. Even though some diseases are terrible, there are also potential risks. People can be harmed by drugs, and we think the best way we can try to predict that is to first put the drug into animals."

Toxicity, he says, may have to be assessed in both a rodent and a nonrodent model, such as a dog or monkey, while some other drug effects may lend themselves to evaluation in species as far from the human as the giant squid.

Katz says the animal experiments don’t even have to prove a treatment is absolutely safe, let alone effective, before the FDA allows researchers to try it in humans. Sometimes, the animal experiments just point investigators toward potential problems so they know what to monitor when they test a drug in people.

Relating a story of a new epilepsy drug that unexpectedly damaged a nerve sheath in the brains of lab animals, he says, "That was a case where you would never have known that there might be something. It isn’t necessarily true that when animals give you a signal, you find it in people. But it’s the only way we have of knowing how to look, and where."

Usually, the FDA works with the drug’s developers to make sure potential problems can be spotted during the human trial. In most cases, Katz says, "the problem is solved in one way or another. You can figure out a way to monitor for the toxicity, or you may find out that the problem is due to a metabolite that the animal makes that the human doesn’t make."

Rarely do animal safety issues stop an experimental treatment from ever being tested in people. "Today’s crazy idea may be tomorrow’s Nobel Prize-winning discovery," Katz says.

Why Not Use a Computer?

While some people want drug testing to proceed directly to people, others urge the use of computers (in silico models) or cells in a dish (in vitro models) to predict human responses, thereby saving time and money and avoiding injury to animals.

Unfortunately, although some scientific questions can be answered by experiments on cells alone or even with computer models, most complex questions require studying a whole organism. No computer or tissue sample can replicate an animal’s metabolic, immunologic or cardiovascular response to a drug, or evaluate a drug’s potential to cause cancer.

Dealing With Differences

There are differences — as well as startling similarities — between, say, rodents and human beings, says Gregory Cox, associate staff scientist at the Jackson Laboratory in Bar Harbor, Maine, that must be dealt with when drugs are tested. (Jackson is a nonprofit genetics research facility that also supplies approximately 2,000 varieties of mice to labs and universities all over the world.)

Compensating for Size and Scale

A frequent criticism of the mdx mouse, a model for human Duchenne muscular dystrophy (see "The Tale of the mdx Mouse,"), is that "outwardly, they really don’t look that sick," Cox says. (The inexperienced researcher who’s testing a treatment in this mouse and expects to see a change in strength before and after it’s given may not see much and may erroneously conclude that the drug doesn’t do anything.)

Cox believes the difference in weakness in human and mouse DMD is related to the scale of the stresses on human and mouse muscles. "Dystrophin does the same thing in mice as in humans," he says. "And the dystrophin-glycoprotein complex [a cluster of proteins in the cell membrane] is the same."

To stress mouse muscles in a way that’s comparable to what human muscles undergo, he says, you have to do things like put the mice on a downhill treadmill — a stressful environment for the leg muscles — and videotape changes in their gait. Additional devices, such as transparent treadmills that allow a scientist to view the mouse from below, and automated gait analysis provided by a computer, can also help.

"The mdx mouse has a huge number of advantages and is quite a good model genetically and biochemically," Cox says. "If you take a muscle biopsy, you can see that the mice have muscular dystrophy. There’s no obvious difference between them and a young muscular dystrophy patient."

Adjusting for Metabolic Rates

"I think some of the biggest differences are in the metabolism of drugs," Cox says, referring to the process by which enzymes in the liver alter a medication’s chemical structure. A drug that may last hours or days in the human bloodstream may only last minutes in mice because of their very high metabolic rate, he says, even though the drug may act the same way in all other respects.

"If the active compound is in the animal at the same level as it is in a person, it usually has the same action," Cox says. "The problem can be getting it to that level." To do that, he says, investigators need to use innovative strategies, such as putting a tiny pump under the mouse’s skin that infuses the drug continuously, or even breeding mice that have human drug-metabolizing enzymes.

Lab mice are exposed to the same environment and food and usually have the same genetic background, making their response to a test drug likely to be the same. Humans, in contrast, live in different environments, ingest a variety of substances, and have diverse genetic backgrounds, so their responses to a test drug are less likely to be uniform. A drug with a very weak benefit may look much better in mice than it does in humans, where background differences may override its effect.

Factoring in 'Reality'

Sometimes drug trials using lab animals are conducted under conditions that are so ideal that they uncover very small, nonmeaningful differences in the treated and untreated groups.

Real-world factors in human trials create far more variation than scientists see in the lab.

With animal model trials, Cox explains, you "control all the variables. None of the mice fail to complete the study. There are no compliance issues and no lifestyle issues."

Mice don’t smoke, drink or take other medications during the study. Even their genetic background can be strictly controlled and kept exactly the same — certainly not the case in human studies.

When treated mice in this kind of trial survive a few days or weeks longer than the untreated mice, "everybody hopes that difference will scale up to years in patients. Unfortunately, that’s often false logic," Cox says.

"Just because something is statistically significant doesn’t necessarily mean it’s clinically relevant," he continues. "You can get statistical significance [showing that a perceived difference is not just due to chance] in a model system by detecting subtle differences and saying they’re important.

"But just being able to say the difference is unlikely to be due to chance doesn’t make it big enough to merit attention."

Cox says people should get excited when researchers find major effects in animal models, using reliable testing methods.

"If you’re hitting a specific disease mechanism, it’s going to have a dramatic impact on the outcome," he says.