"In the range from a normal breath in to a forceful breath in, we don't have a lot of variation in the amount of insulin delivered," said Alkermes's Silverman. "It's more like swallowing a pill with water. You need a little water to get the pill down. But even if you drink a gallon more after that, it doesn't matter. The pill is already in your stomach."

The Mannkind inhaler also is just larger than the palm of a hand in the closed position, but nearly doubles in length when the mouthpiece unfolds.

The Kos inhaler, like the asthma inhalers it was modeled on, works with a propellant gas. The insulin, mixed with just water and gas, never requires refrigeration. Unlike other inhalers, which must be loaded with a dose of insulin from a capsule or blister pack every time they are used, the Kos inhaler has a reservoir of 120 "puffs" of insulin. A puff counter calibrates dosing and keeps track of the remaining supply of insulin on a small digital display. If a patient knows he needs about six or seven puffs a day, he can check the puff counter to determine how long his supply will last.

All of the companies benefit when Pfizer, the mega-marketer, begins selling doctors and patients on the new approach to therapy. But developers at the four companies with drugs in clinical trials also are obliged to answer questions that, so far, have been raised only by Pfizer's product. Each of the other companies, for instance, must build into their Phase-III clinical trials a two-year safety study. Not because they have shown any negative data, but because Pfizer and Nektar's drug, which is formulated differently from their own, was linked to a slight and temporary decrease in lung function. Pfizer has agreed to begin a five-year post-launch surveillance study of 5,000 patients that is expected to confirm what the FDA advisory panel already documented: a temporary decrease in lung function that stabilized within six months, did not progress, and was reversible when the drug was discontinued. Ironically, the extra study required by FDA extends the time that Exubera has sole possession of the market.

"Right now they're telling everybody who wants to develop an inhaled insulin, you have to have two years' data" says Rosskamp. "Because right now they're assuming that everybody will see this lung function decline, which Exubera has seen in their clinical trials." FDA requires Kos to begin with a certain number of patients (which it declines to reveal) for the initial six-month study and to retain some of them for a year, and a final group for another year. FDA officials declined to be interviewed for this article.

Kos and, less explicitly, Mannkind, have indicated that they believe this extended safety trial may not prove necessary for their drug. Each company hopes that its formulation, which has no excipients, will prove less irritating to the lungs than Exubera. Rosskamp made the case most directly: "One of the strategies which we try to pursue with the FDA is, if we show that after six or 12 months we do not observe this decline, then we wouldn't need two years' data. So we could file earlier. That's something, of course, which we'll have to discuss with them."

However the safety study works out, none of the new products is likely to reach the market before 2008. And if they must complete the full two-year trial, manufacturers may not launch until 2009 or 2010. Until then, Exubera's performance in the marketplace is likely to matter more than the drug candidates' results in clinical trials.

"Exubera has much to prove on the inhaled-insulin market," says Wood Mackenzie's Kidd. "Not only for Pfizer, but also for some of the guys that are hoping to be fast followers. If Exubera falls flat on its face, they will face a very tough challenge to reinvigorate the market and gain patient and physician confidence again."