Case Study 2 — Inhaled Insulin, Twice

A delivery problem solved technically and failed commercially — and then failed again

Type: Real, public, commercial failure · Tier 1 approval facts, Tier 2 commercial detail · Relevance: §4.1, §4.4, §4.6

Background: the needle problem is real

People with diabetes inject insulin, often several times a day, for decades. The burden is genuine — not merely unpleasant but a documented contributor to delayed initiation of insulin therapy and to poor adherence once started. Clinicians describe patients resisting insulin for years, accepting worse glycemic control rather than begin injecting.

So a needle-free insulin has been one of pharmaceutical development's most obviously valuable targets for a century. Oral delivery is essentially impossible for insulin (§4.1). But the lung is interesting: an enormous surface area, extremely thin alveolar membranes, dense blood supply, and relatively low protease activity compared with the gut.

Two serious attempts were made. Both worked pharmacologically. Both failed.

Attempt one: Exubera

Exubera, an inhaled insulin powder, was approved by the FDA in 2006 after a full development program. It was not a marginal approval — it worked. Insulin delivered to the lung reached the circulation and lowered blood glucose.

It was withdrawn from the market in 2007, roughly a year later, having achieved very poor sales. The manufacturer took a substantial write-off.

The failure was not one thing. It was several, and each of them maps onto something in Chapter 4.

The device was large. The inhaler was frequently described as roughly the size of a tennis ball can. A patient could now avoid an injection they could perform discreetly with a small pen, in exchange for using a conspicuous device.

Dosing was in unfamiliar units. Insulin dosing is done in units, and patients and clinicians think fluently in them. The inhaled product came in fixed-size blisters that mapped awkwardly onto that mental model. Dose adjustment — the daily work of insulin therapy — became harder rather than easier.

Lung function monitoring was required. Because the drug was being deposited in the lung chronically, spirometry was recommended before starting and periodically thereafter. A therapy that was supposed to reduce burden added a monitoring requirement.

And smokers were excluded, because smoking substantially increases pulmonary insulin absorption and makes it unpredictable — a variability problem of exactly the kind §4.4 identifies as the real killer.

Attempt two: Afrezza

A second inhaled insulin, Afrezza, was approved in 2014. The design addressed several of Exubera's specific failures: a much smaller, more discreet inhaler, and a formulation with a notably rapid onset — arguably faster than injected rapid-acting insulin, which is a genuine pharmacological advantage for controlling the glucose rise after a meal.

Commercially, it has remained a very small product. It carries a boxed warning regarding bronchospasm in patients with chronic lung disease, requires lung function assessment, and is contraindicated in asthma and COPD. It remains available, and it has never approached the market position that a needle-free insulin was expected to command.

Two well-funded attempts, both technically successful, both commercially marginal.

🩺 Safety and Risk — why the lung is a harder route than it looks

The lung's attraction is its surface area and thin membranes. Its problems are less obvious and mostly not about absorption at all.

The lung is not a disposal organ. Anything deposited there stays in contact with delicate tissue. A drug given for decades, several times daily, is a chronic pulmonary exposure — and the safety question is not "does the dose reach the blood" but "what does the residue do to the lung over thirty years?" That question cannot be answered by a two-year trial, which is why lung function monitoring accompanied both products.

Absorption depends on the state of the lung. Smoking increases it substantially. Respiratory infection changes it. Asthma and COPD change it. A patient with a cold may absorb a different fraction of their dose than the same patient the week before — and for insulin, where dosing errors produce hypoglycemia, that variability is dangerous rather than merely inconvenient.

And particle size determines destination. Too large and the powder deposits in the mouth and throat; too small and it is exhaled. The usable window is narrow and requires precise engineering of both formulation and device.

This is a general lesson about routes. §4.4's table gives each route a bioavailability figure, which makes them look comparable. They are not. Each route has its own failure modes, its own variability profile, and its own long-term exposure question, and the bioavailability number captures none of it.

What this case teaches

Technical success and therapeutic success are different things. Both products did what they were designed to do. Insulin reached the bloodstream via the lung and lowered glucose. Judged as pharmacology, both were achievements.

The delivery problem is never only about the molecule. Chapter 4 frames delivery as a biochemical challenge — enzymes, barriers, clearance. These cases add the rest of it: the device, the mental model the patient uses to dose, the monitoring burden, the excluded populations, the cost, and what the therapy asks of someone for the next thirty years. A route that solves the biochemistry and loses on the rest has not solved the problem.

Variability defeats otherwise adequate routes. Both products' most serious pharmacological liability was that absorption depended on lung condition. §4.4 states that consistency matters more than magnitude; this is the case that proves it, in a therapy where the consequence of an unexpected dose is hypoglycemia.

And a good idea can fail twice without being wrong. Needle-free insulin remains a worthwhile goal. Neither failure demonstrates that the concept is unsound — they demonstrate that two specific implementations lost on specific grounds, several of which were fixable and some of which were fixed the second time. Whether a third attempt is worth making is a genuinely open commercial and scientific question.

Compare with oral semaglutide, which took a route with roughly 1% bioavailability and burdensome administration requirements and succeeded anyway. Why? Partly because the therapy is once daily rather than several times a day, partly because the consequence of a missed or reduced dose is much less acute than for insulin, and partly because the alternative it competes with is a weekly injection rather than a discreet pen used at every meal. The same delivery compromise is acceptable in one context and not in another, and nothing about the pharmacokinetics tells you which.

Discussion questions

1. Both inhaled insulins worked pharmacologically and failed commercially. Is "the drug worked" a meaningful statement about a product nobody uses? What is the right unit of success?

2. Exubera required lung function monitoring; the therapy was supposed to reduce burden. Identify at least three other ways a treatment intended to reduce burden can end up increasing it.

3. §4.4 argues that variability is more damaging than low bioavailability. Explain why, using insulin specifically — what is the consequence of an unpredictable dose here, versus for a weekly GLP-1 agonist?

4. Afrezza fixed several of Exubera's specific problems and remained a small product. What does that suggest about whether the failures were correctly diagnosed? Propose an alternative diagnosis.

5. Oral semaglutide accepted ~1% bioavailability and strict administration requirements and succeeded. Inhaled insulin accepted a device and monitoring requirement and did not. Identify the specific features of each therapy that made the same category of compromise acceptable in one case and not the other.

6. Apply the delivery filter (§4.9) to inhaled insulin. Which of the five questions does it pass? Which would you have to add to the filter to have predicted these outcomes?