MannKind announces publication of STAT trial (Afrezza vs. NovoLog), with updated efficacy results, in DT&T – September 13, 2018

Executive Highlights

  • MannKind announced today that full results from the STAT study have been published in Diabetes Technology & Therapeutics. Several updates to the topline results presented at ADA 2018 are included in the paper, including an increase in the reported time-in-range benefit of Afrezza vs. NovoLog from 1.5 to 2 hours/day – it sounds like statistical analysis the first time around was conservative. Time in hypoglycemia and hyperglycemia was also shared, significantly favoring Afrezza. Presumably, the difference arises from definitions of “compliance”; only 15 of 26 participants randomized to Afrezza were included in this analysis.

  • In an exclusive interview with our team, MannKind CEO Dr. Michael Castagna shared that a pediatric study of Afrezza is set to begin in 2019. A pediatric study would be helpful for the product’s uptake, which to date has been slow due to reasons very unrelated to demand, from our view.

  • MannKind will have three posters and one oral presentation at EASD. According to management, two of the posters and the oral presentation will be devoted to STAT, while the final poster will examine respiratory outcomes of patients on the inhaled insulin.

MannKind announced yesterday that full results from the STAT study have been published in Diabetes Technology & Therapeutics. STAT (n=60) compared MannKind’s ultra-rapid-acting inhalable insulin Afrezza (Technosphere insulin) to Novo Nordisk’s NovoLog (insulin aspart); Dexcom G5 was used over the four-week study.

We see this publication as a victory for MannKind and Afrezza. In our observation, the vast majority of physicians are not fully aware of Afrezza’s benefits over other mealtime insulins, beyond its inhalable nature (if that), and publication of these results should improve Afrezza’s visibility. Moreover, enthusiasm is very strong for the product among thought leaders (including Drs. Irl HirschBruce Bode, and Satish Garg); as well, MannKind has faced significant obstacles in securing reimbursement for the product, keeping it out of reach for many. Although a number of patients we know on Afrezza would say it is “faster” and that it feels like it has less hypoglycemia associated with it, this is not proven and not yet part of the label. Nonetheless, we hope that publication of STAT will help bring more prescribers on board and that MannKind can leverage Afrezza’s clinical profile into better coverage.

STAT CGM-Based Outcomes (LS means ± SE; adjusted for age, sex, study site, and total insulin dose). Patients were defined as “compliant” to the Afrezza regimen if at least 90% of post-meal doses were taken per protocol, with at least one of the post-meal inhalations taken if indicated per meal. We wish they’d use the word adherent! Yellow = statistically significant (p<0.05) compared to NovoLog; Blue = statistically significant (p<0.05) Afrezza noncompliant compared with Afrezza compliant group. Percent values are of a 24-hour day. It is a bit unusual to look at data in terms of who is compliant and who is not in such a small study – we also think it’s a little odd to have 34 people with diabetes on Novolog and then 22 on Afrezza, with just under 70% labeled “compliant.”


Aspart (n=34)

Afrezza Noncompliant (n=7)

Afrezza Compliant (n=15)

% Time in Range (70-180 mg/dL)




% Time in Hyperglycemia (>180 mg/dL)




% Time in Hypoglycemia <70 mg/dL




% Time in Hypoglycemia <60 mg/dL




% Time in Hypoglycemia <50 mg/dL




Revisions to Topline Results

  • The publication includes several considerable revisions to the topline readout from ADA, in the direction of being more favorable for Afrezza but for a smaller group. A significant (p<0.05) 2.1 hour increase in time-in-range was found in patients who complied to their Afrezza regimen compared to patients on NovoLog – up from 1.5 hours in the initial readout (p<0.005). Given that we saw 1.5 additional hours of time-in-range per day as a very meaningful outcome for patient quality of life, we’re positive about the update, though it’s relevant for a smaller number of patients, which we are trying to better understand. At the topline results presentation, the highly-respected Dr. Satish Garg spoke highly of Afrezza as filling an unmet need in diabetes care for faster mealtime insulin that better addresses postprandial glycemia, and he attributed these positive findings to Afrezza’s very rapid onset (~12 minutes to first measurable effect) and offset (~90-180 minutes to return to baseline, depending on dose). It’s our sense that this prompts less insulin stacking, which is associated with less hypoglycemia (though again this is not on the label).

  • In the initial readout, postprandial glucose excursions (defined as the peak increase in CGM glucose in the first four hours after each meal), were found to be 15% lower (-20 mg/dl) in those taking Afrezza, regardless of compliance and compared to NovoLog. The publication indicates that only those who were compliant to the Afrezza regimen experienced a significant decrease in postprandial excursions (p<0.05) compared to NovoLog, though no exact value was given (Figure 3B).

  • Study size was reduced from 60 in the initial readout down to 56 in the publication analysis, with all four subjects removed from the Afrezza group. Presumably, these removals account for the revisions to the outcomes above. We are trying to better understand the removal of these subjects from the analysis.

New Outcomes

  • Significantly less time was spent per day in hyperglycemia (>180 mg/dl) by those who were adherent to their Afrezza regimen compared to patients on NovoLog (8.2 hours/day vs. 9.8 hours, p<0.05). This said, the most time was spent in hyperglycemia by those who were noncompliant to Afrezza, and the difference between the compliant and noncompliant groups was also statistically significant. This difference was also reflected in the time-in-range data for each of the groups, for which the noncompliant and NovoLog groups were virtually identical, while the compliant group did significantly better. All of the time-in-range benefit was seen during daytime (6 am to midnight), which likely reflects that the benefit is coming through improvements in postprandial glucose. While we don’t like the phrases compliant and noncompliant, we are keeping them for ease of use for this report.

    • As we understand it, the compliant/noncompliant breakdown of the results was undertaken to account for a dosing issue in the design of STAT. Importantly, the study was designed prior to the proper dosing conversion from NovoLog to Afrezza (1 unit:1.5 units) was “discovered.” In STAT, pre-meal insulin was converted 1:1, meaning that a post-meal dose of Afrezza was often necessary to ensure an equal amount of total insulin-on-board between the NovoLog and Afrezza groups. Accordingly, in an exclusive interview with our team, MannKind management attributed the significant difference in hyperglycemia and time-in-range between the noncompliant and compliant groups to this improper dosing conversion. Moreover, they emphasized that they would expect both better outcomes and a reduced need for post-meal doses of Afrezza with a 1:1.5 conversion, which will be used in all future studies.

  • On average, Afrezza users spent numerically less time in hypoglycemia <70, <60, and <50 mg/dl, though this difference was not significant when comparing either the non-compliant or compliant. The difference amounts to ~nine fewer minutes <50 mg/dl and ~24 fewer minutes <70 mg/dl with Afrezza vs. NovoLog, based on the breakdown in the table above. More positively, however, a significant reduction in time spent in hypoglycemia <60 mg/dl (p=0.02) and <50 mg/dl (p=0.04) was found when comparing the combined Afrezza groups to NovoLog (Table 4), when adjusting for study week, A1c at screening, and the variables accounted for in the table above. In our opinion, this seems a somewhat convoluted analysis, but we also think the data clearly trend in favor of a hypoglycemia benefit with Afrezza.

Future Studies

  • Several references to future clinical trials for Afrezza were made in the publication (Dr. Garg described STAT as a pilot at ADA). MannKind CEO Dr. Michael Castagna shared the company’s plans in an exclusive interview with our team. We’ve heard anecdotally that hypoglycemia virtually disappears with Afrezza, and we would love to see this deeply important benefit reflected on the label assuming this is the ‘broad’ experience, which we think it is. A revamped STAT study (12-24 weeks, 1:1.5 pre-meal conversion of NovoLog to Afrezza) is also in the works, though no timeline was provided.

  • MannKind will have three posters and one oral presentation at EASD. According to management, two of the posters and the oral presentation will be devoted to STAT, while the final poster will examine respiratory outcomes of patients on the inhaled insulin.

  • We got to speak with a range of leaders at the company including new Chief Marketing Officer Garrett Ingram, who was incredibly impressive in our conversation – MannKind is very lucky to get her leadership!

Close Concerns Questions

Although hyperglycemia was significantly higher in the noncompliant Afrezza group than the compliant group, the two-hour post-meal AUC was still improved regardless of compliance (Figure 4A). When was the hyperglycemia experienced during the day, if not post-meal?

Dinner was the only meal in which a significant difference in postprandial glucose excursions was not observed (Figure 3D), though there was no breakdown by compliant vs. noncompliant. Management noted that much of the noncompliant group likely comes from the subset of people who (understandably) did not take a post-dinner Afrezza dose due to fear of overnight hypoglycemia. What does the figure look like when these groups are separated?

Why were four Afrezza-randomized patients removed from the analysis? Were they in the compliant or noncompliant Afrezza group? We acknowledge that there are valid reasons to remove data points, but we would love to know the reasoning behind this change.

Why were different variables adjusted for when comparing the NovoLog group to all Afrezza users (Table 4) as opposed to the compliant and noncompliant groups separately (Table 5)? Did these adjustments confer significance where otherwise it would have been absent?


--by Peter Rentzepis, Ann Carracher, and Kelly Close