Oral 11β-HSD1 inhibitor AZD4017 improves wound healing and skin integrity in adults with type 2 diabetes mellitus: a pilot randomized controlled trial

Background Chronic wounds (e.g. diabetic foot ulcers) reduce the quality of life, yet treatments remain limited. Glucocorticoids (activated by the enzyme 11β-hydroxysteroid dehydrogenase type 1, 11β-HSD1) impair wound healing. Objectives Efficacy, safety, and feasibility of 11β-HSD1 inhibition for skin function and wound healing. Design Investigator-initiated, double-blind, randomized, placebo-controlled, parallel-group phase 2b pilot trial. Methods Single-center secondary care setting. Adults with type 2 diabetes mellitus without foot ulcers were administered 400 mg oral 11β-HSD1 inhibitor AZD4017 (n = 14) or placebo (n = 14) bi-daily for 35 days. Participants underwent 3-mm full-thickness punch skin biopsies at baseline and on day 28; wound healing was monitored after 2 and 7 days. Computer-generated 1:1 randomization was pharmacy-administered. Analysis was descriptive and focused on CI estimation. Of the 36 participants screened, 28 were randomized. Results Exploratory proof-of-concept efficacy analysis suggested AZD4017 did not inhibit 24-h ex vivoskin 11β-HSD1 activity (primary outcome; difference in percentage conversion per 24 h 1.1% (90% CI: −3.4 to 5.5) but reduced systemic 11β-HSD1 activity by 87% (69–104%). Wound diameter was 34% (7–63%) smaller with AZD4017 at day 2, and 48% (12–85%) smaller after repeat wounding at day 30. AZD4017 improved epidermal integrity but modestly impaired barrier function. Minimal adverse events were comparable to placebo. Recruitment rate, retention, and data completeness were 2.9/month, 27/28, and 95.3%, respectively. Conclusion A phase 2 trial is feasible, and preliminary proof-of-concept data suggests AZD4017 warrants further investigation in conditions of delayed healing, for example in diabetic foot ulcers. Significance statement Stress hormone activation by the enzyme 11β-HSD type 1 impairs skin function (e.g. integrity) and delays wound healing in animal models of diabetes, but effects in human skin were previously unknown. Skin function was evaluated in response to treatment with a 11β-HSD type 1 inhibitor (AZD4017), or placebo, in people with type 2 diabetes. Importantly, AZD4017 was safe and well tolerated. This first-in-human randomized, controlled, clinical trial found novel evidence that 11β-HSD type 1 regulates skin function in humans, including improved wound healing, epidermal integrity, and increased water loss. Results warrant further studies in conditions of impaired wound healing, for example, diabetic foot ulcers to evaluate 11β-HSD type 1 as a novel therapeutic target forchronic wounds.


Secondary outcome analysis
The secondary endpoints, systemic 11β-HSD1 activity and skin function, and continuous clinical laboratory safety variables were analyzed in the same ways as the primary endpoints.
Adjusted summaries were obtained using a linear regression model that mirrored the analysis of covariance approach for wound healing and laboratory safety variables. For all other variables, visual inspection of linear regression model residuals indicated that they were not normally distributed. For TEWL and epidermal integrity, log-transformation was performed before linear regression. For the remaining outcomes, for which log-transformation did not render normally-distributed linear regression model residuals, quantile (median) regression was used.
Planned supplementary analyses used linear mixed modelling to allow the precise timing of measurements to be included as a covariate where relevant and to account for any differences in timings between groups. Likelihood ratio tests supported the inclusion of non-linear (quadratic) terms for change over time and supported allowance of changes over time to vary between patients.
The standalone SAP pre-specified that in the event that there was not proof-of-concept for the primary outcome, effects on secondary outcomes would be interpreted with reference to the measured effects on systemic, as opposed to skin, 11β-HSD1 activity. If both showed a potential difference between groups, this would be considered preliminary evidence of efficacy. However, without evidence of a substantive difference in 11β-HSD1 activity either systemically or in the skin, any apparent substantive differences in secondary outcomes were to be interpreted with caution.

Additional pre-planned analyses
Correlations between plasma AZD4017 concentration at day 35 and skin AZD4017 concentration at day 28 and between AZD4017 compliance and efficacy outcomes in the active treatment arm were estimated using Spearman rank correlation.
We measured the strengths of associations between skin 11β-HSD1 activity and skin outcome measures while controlling for systemic GC level using partial correlation after rank transformation. Correlation coefficients were transformed using Fisher's z transformation before averaging across multiple imputed datasets. For all correlation analyses, absolute correlation coefficients with a value of r(ho) greater than 0.3 were considered preliminary evidence of substantive association.
The numbers of patients with clinical laboratory values below, within, or above normal ranges before the intervention and at each post-intervention time point were tabulated for each test for the safety population by treatment group. The proportions of patients who passed the overall assessment of blood safety at days 0, 7, 28, 35, 42 were summarized.
For AEs, summaries of incidence rates (frequencies and percentages), intensity, and relationship to study drug of individual AEs by system organ class and preferred term are presented.
Feasibility variables were summarized descriptively.
Estimates of sample sizes for future trials were based on the pooled SDs from both treatment arms for the following outcomes: 11β-HSD1 activity in skin (at 28 days); sudomotor function, skin hydration, epidermal barrier function, integrity (at 28 days), and recovery (TEWL at 3 hours, 2 days, and 7 days after disruption by repeat tape stripping at 28 days); skin thickness (at 35 days); and WH (maximal early granulation tissue width at 2 days and maximal blood clot depth at 7 days after biopsy at 28 days).

Recruitment halted early
Study recruitment was intended to continue until a total of 30 participants had been randomized to ensure recruitment of at least 12 per group, as recommended for pilot studies, with a 20% dropout rate. The recruitment period was extended twice. At the end of the second extension, 28 patients had been recruited, and the dropout rate, at less than 5%, was lower than expected. Therefore, we decided to halt recruitment rather than extend it further, which would have increased the risk of delaying the reporting of the trial results and the planning of future trials based on these results. This decision was made without reference to the primary outcome measure, whose results had not yet been processed, and before the breaking of the blind.

Primary outcome unit of measurement
After the completion of the study and final database lock, the primary outcome was found to have been calculated as percent conversion per 24 hours, rather than per hour as stipulated in the protocol. As this difference in scale affected all values equally and did not affect the conclusions, the sponsor and investigators agreed that the values would not be changed and would be reported as percent conversion per 24 hours.

Validation of primary outcome by ELISA
In a protocol amendment, validation of the radioimmunoassay method of measuring 11β-HSD1 activity in the skin by a cortisol ELISA was added. This addition was made before the processing of the biopsy samples and the breaking of the blind.

Measurement of wound depth instead of diameter at days 7 and 35
At 2 days after wounding, maximal early granulation tissue width (a marker of early healing) was prespecified as the standardized indicator of wound diameter. However, the tissue width had reduced to zero in all patients by 7 days after wounding. Therefore, maximal clot depth (a 5 marker of later healing that was absent at 2 days after wounding) was substituted as the standardized indicator of healing at this time point

Per-protocol analysis
Only one participant who had withdrawn from follow-up due to work commitments after day 7 was excluded from the per-protocol dataset. Because the per-protocol analysis of each outcome was designed to include only participants in the per-protocol population who had data available for that outcome, the per-protocol analysis was essentially identical to the planned available case sensitivity analysis.

Additional sensitivity analysis of multiply imputed data
An additional planned sensitivity analysis that would have increased or decreased imputed values in multiples of the baseline SD in the observed data was not performed because of the low level of missing data.

Safety population
The safety population or safety set includes all participants who received any amount of the planned study medication.

Efficacy population
The efficacy population or full analysis set includes all participants who were randomized and received at least one dose of the planned study medication.

Per-protocol efficacy population
The per-protocol efficacy population includes all participants in the efficacy population, except for those who met the following criteria: 6 − Receipt of prohibited prior, concomitant, or prior and concomitant medications − Failure to meet inclusion or exclusion criteria (i.e., those who entered the study in error) − Overall compliance with study treatment during the trial less than 80% − No receipt of study treatment to which they were assigned through randomization − Withdrawal from study treatment for any reason For each primary and secondary variable, at each visit, only participants in the per protocol population with data available were to be included in the per protocol analysis.