Diabetic Foot Ulcer Study on Topical Interventions

Patient flow The target subjects were screened in the orthopedic department of two regional hospitals and one general outpatient clinic (GOPD). Those eligible potential subjects were recruited after they discharged from hospital according to the selection criteria. All the subjects were referred and intervened in an orthopedic nurse clinic a regional hospital. The subjects were randomized into three groups through a fixed block of 10 by online software. A sequence of assignment was generated and put into a sealed envelope by a research assistant. The investigator did not involve in randomization process and allocation sequence. The subjects were enrolled and allocated to different treatment arms according to the allocation sequence.

Data collection Outcome assessor blinding was employed in this study. A research assistant was responsible for measuring the wound size and taking clinical photo. In every clinical visit, the research assistant waited outside the clinic until the removal and proper cleansing of the wound by the first author so that she unaware of the topical treatment option. The subjects would be discontinued follow up when the wound was completely healed or till the end of 12-week study period.

Intervention All participants attended the nurse clinic for follow-up by the first author (nurse consultant) weekly in the first four week and then biweekly till 12 weeks of follow up period. They were nine clinical attendances totally. At each visit, sharp debridement for the non-viable tissue and stimulation of the vascularity on the avascular tissue were performed if needed by the first author. Then, he applied the topical dressing according to randomization sequence.

Statistical analysis All the analysis will be carried out according to the intention-to-treat principle. SPSS Statistics for Mac version 22 (SPSS Inc, Chicago, Illinois) was used for data analysis. Comparison would be made among groups by Fisher's exact test for nominal data and Kruskal-Wallis test for ordinal and scale data. The complete ulcer healing was compared among groups by Kaplan-Meier estimates. General estimating equation (GEE) was used to compare the ulcer size reduction rate, the wound fluid concentration of matrix metalloproteinase -9 (MMP9), tumor necrosis factor alpha (TNF-α) and interleukin-1 alpha (IL-1α) among groups. Statistical significance was set at p < 0.05 for all tests.

Sample size calculation

• The probability of non-healing (p1) on nanocrystalline silver (nAg) group = 0.20
• The probability (p2) of non-healing on manuka honey (MH)/conventional group = 0.50
• The alpha (α) value was 0.05 and power (1-β) was 0.8
• Ratio of sample size per group was 1:1
• Therefore, the total sample subjects for 2 groups were 77. Each group needed 39 subjects. For 3 groups, the total numbers of subjects were 117.
• From the experience, 10% of patients were loss to follow up. The actual numbers of subjects needs were 129. i.e. 43 subjects per group.

Data quality assurance In order to ensure the data quality, a research assistant was assigned to verify the source document and other trial records were accurate, complete, kept up-to-date and maintained. The student investigator trained the research assistant in the data quality assurance process. The student investigator also checked the data entry sheet with raw data every month. The responsibilities of research assistant were listed as follows.

• Check the accuracy and completeness on the case screening form and consent form.
• Verify the laboratory data in the case record form (CRF) were consistent with the corresponding laboratory result.
• Verify the inter-current illness was reported on the CRF.
• Verify all withdrawals and dropouts of enrolled subjects from this trial were reported and explained on the CRF.

If there was any missing data or data inconsistency, the research assistant was clarified the data with the internal electronic record of the hospital.

Handling of missing data The nature of this study was the multiple observations of subjects in regular censored points. Thus, the missing clinical data was handled by the last observation carried forward. For those laboratory data, statistical method of general estimating equation was used for the analysis on the repeated measures among groups. The missing data was under the missing completely at random (MCAR) assumption. Therefore, the laboratory missing data did not need to make estimation in the present study.