Setting. The study was conducted at two hospital-based wound care centers in three phases. An onsite monitor was present to observe the collection of data throughout the study.
Patient eligibility and randomization. The first phase (Screening) consisted of a series of screening assessments designed to determine study eligibility (see Table 1). Baseline measurements were taken and recorded before study enrollment. Patients who met the study eligibility criteria were randomly assigned to one of two treatment cohorts. Randomization was performed using a series of sealed envelopes that designated the biologically active treatment to be applied. Envelopes were randomized in blocks of six; however, the investigators were unaware of the block size or randomization scheme.
Block randomization is a widely used technique to prevent the investigator from second-guessing the randomization scheme. In this case, with a block of six, three out of every six envelopes would contain assignment to one treatment and three to the other. If the investigator was aware of the block size, he/she would know if three assignments to group one had been made; the remaining three assignments would be to the alternate group. Because the investigator did not know the block size, he/she still could not guess what the next envelope contained. Envelopes were filled by a third party using a randomization table or similar technique. Blocks are necessary to keep the group assignments approximately balanced. Although unlikely, without blocks, it would be possible to assign 20 participants in a row to the same treatment, depending on how the assignments were drawn. Block assignments guarantee that once six subjects have been enrolled, three will be placed in each group.
Patient demographic and wound characteristic variables recorded in the case report forms and included: age, BMI, diabetes type, race, ethnicity, gender, duration of diabetes, presence of insulin dependence, number of diabetic medications the patient currently was taking, wound duration, presence of peripheral artery disease (PAD), presence of peripheral neuropathy, smoking status, baseline wound size, and number of grafts used. Patient anonymity was maintained by assigning study code numbers to each patient at the beginning of the study; these codes were used as identifiers throughout the period of time during and after the participant was enrolled in the study. Codes could only be broken by the principal investigator in the event of an emergency.
Once randomized, patients returned 1 week later to begin treatment. Patients who presented with more than one wound were permitted to participate in the study. In this case, the study wound was the largest wound, and all edges of the study wound had to be at least 2 cm from the nearest adjacent wound edge.
Treatment and evaluation. In the second phase (Treatment), the patient returned to the clinic for initial application of the biologically active product. Patients randomized to the HSA group received a product application every other week; those randomized to the HFDS group were treated every week in accordance with each manufacturer’s product instructions.9,13 In addition, preparation of the product (ie, thawing) before application was done precisely according to manufacturer instructions. In each case, wounds were prepared in the same way: nonviable tissue and callous were debrided from the wound surface and surrounding wound perimeter, respectively. Wounds were thoroughly cleansed with saline before application of the treatment product. The investigator physician at each site performed wound debridement and applied the biologically active product. All investigators were trained by company representatives for each product in order to ensure correct technique and had attended an investigator meeting to guarantee uniformity in technique across sites.
Once applied, both study products were covered with either Mepitel (Mölnlycke Health Care US, LLC, Norcross, GA) or Polymem (Ferris Manufacturing Corporation, Fort Worth, TX), according to the investigator’s preference. All wounds were offloaded with ½-inch felt as part of an aperture type of device. All participants were given either a healing sandal fabricated from a surgical shoe (Darco, International, Inc, Huntington, WV) or a fixed ankle boot (DH Off-Loading Walker; Ossur, Foot Hill Ranch, CA) if the patient required more protection. The decision to treat with either a surgical shoe or fixed ankle boot was based on the clinician’s assessment of the wound and the patient’s need for more stability. One participant in each group received the fixed ankle boot, and one in the HFDS group was treated with a total contact cast due to the inability to tolerate either a healing shoe or fixed ankle boot.
Treatment evaluations included investigator assessment and documentation on data collection sheets of wound closure and wound size. Only fully epithelialized wounds were considered healed. Because the grafts have a different physical appearance, it was not possible to disguise the type of graft used at the time of evaluation. At each visit, wounds were photographed and the margins of the wounds were traced along the inner edge of the margins. This was done after removal of the calloused perimeter, when present. Wound area was determined by measuring the length and width of the wound from the tracing and inserting these measurements into the equation for an ellipse. This is the same technique as described by Gilman.16 Adverse events, new medications, and any changes in the medical history were recorded. Any evidence of infection, as well as associated treatments, also was documented. No additional biologically active materials could be used and no vascular interventions were permitted during the course of the study. Patients who received either of these types of treatments during the course of the study were excluded from this analysis. The treatment phase continued for 12 weeks. However, if the wound closed before 12 weeks, one follow-up visit was performed to confirm wound closure. The time to closure was measured based on the visit where closure was first observed, not on the confirmation visit.
Follow-up. Patients with an unhealed wound at week 12 continued on to the third phase (Follow-Up) for treatment and evaluation for up to eight additional weeks. No additional biologically active products were used in either treatment group after the week 12 visit. Patients who still had ulcers were treated with saline-moistened gauze and debridement as needed, per the investigators. Persons whose ulcers were recorded as healed (100% closure as determined by the investigator) at the last scheduled treatment visit returned for one confirmatory visit. Participants with incomplete wound closure continued to be assessed until completion of week 20, at which time any subsequent treatment was provided outside of the scope of the study.
Data and statistical analysis. All data were transferred from the data collection sheets to a digital spreadsheet for analysis by an experienced, professional research monitor. The correctness of the data and any missing data points were noted. Each investigator confirmed the correct transfer and interpretation of their data. Unless specified otherwise, SAS Version 9.0 (SAS Institute Inc, Cary, NC) or higher was utilized to perform all statistical analyses. Continuous variables were summarized using general descriptive statistics, and categorical variables were summarized by percentages for proportions. Change in wound size per wound measurements was based on a comparison to the baseline area.
Data were analyzed using Student’s t-test and chi-square analysis to determine the equivalency of the two cohorts. Continuous covariates were assessed for normality and dichotomized or broken into quartiles if normality could not be justified (eg, BMI, wound duration, and baseline wound size were broken into quartiles for the analysis). Normality assessment included the Shapiro-Wilks test for normality, stem leaf plot, box plot, and normal probability plot. Any variable containing >20% missing data was excluded from further analysis. Any categorical or binary variable identified to cause a possible nonconvergence problem from pure results was collapsed if nominal or treated as a continuous variable if ordinal.
Multivariate logistic regression was used to evaluate the association of biologically active therapy agent and the proportion of healed wounds after 12 weeks from initial allograft application in order to control for identified confounders. Hypothesis testing was performed at the 5% significance level, with two-sided P values rounded to two decimal places. All confidence intervals had 95% coverage.
A Kaplan-Meier curve also was generated to demonstrate the cumulative percentage of healing in both groups.
Ethical considerations. Written informed consent was obtained from all study participants. This clinical research study was approved through the Western Institutional Review Board, Puyallup, WA.