Cellular Versus Acellular Matrix Products for Diabetic Foot Ulcer Treatment: The Dermagraft and Oasis Longitudinal Comparative Efficacy Study (DOLCE)—A Randomized Clinical Trial Free

Chronic diabetic foot ulcers (DFUs) affect ∼40 to 60 million people worldwide and 1.6 million Americans each year (1,2). In addition to multiorgan comorbidities, decreased quality of life, and increased hospitalization duration in patients with diabetes, ∼20% of moderate to severe DFU infections result in amputation (2,3).

The estimated 2017 U.S. expenditure for DFU treatment and complications was $80 billion, approximating the $80.2 billion spent on cancer in 2015 (4,5). Despite these costs, with current standard of care (SOC), only 35–59% of ulcers heal within 12 weeks (6–9). This suboptimal success rate has prompted alternative approaches, including those targeting the abnormal extracellular matrix (ECM) by providing exogenous ECM components derived from cellular sources without viable cells, such as cadaver, amniotic membrane, porcine, or bovine collagen (10). However, existing data from randomized controlled trials (RCTs) suggest few can increase the percentage of DFUs that heal by 12 weeks to >50% (11–15). Cost also varies greatly, from $110 per application for acellular matrix (ACM) products to $1,800 per application for cellular matrix (CM) products (16).

Dermagraft, a CM containing living human neonatal foreskin fibroblasts cultured on bioabsorbable polyglactin mesh, is U.S. Food and Drug Administration (FDA) approved for up to eight applications for nonhealing full-thickness DFUs. Several RCTs have demonstrated its ability to improve DFU healing relative to control treatments. In initial trials, 39% of patients using Dermagraft healed at 12 weeks compared with 32% of those receiving SOC. In subsets of patients who received the most viable cells, 51% of ulcers healed with Dermagraft compared with 32% with SOC (17). In the second trial, 30% of Dermagraft-treated DFUs healed compared with 18% of those treated with SOC (11).

Oasis, an ACM scaffold derived from porcine small intestinal submucosa, is FDA cleared for treating DFUs. In one trial comparing Oasis (n = 37) with Regranex gel (n = 36; becaplermin, a recombinant human platelet-derived growth factor), 49% of Oasis-treated wounds healed compared with 28% of Regranex-treated wounds (15). Another trial (n = 13 per arm) compared eight applications of Oasis with three applications of Dermagraft over 12 weeks and revealed 77% of DFUs had healed with Oasis versus 85% with Dermagraft by week 20. However, this trial was underpowered and lacked blinding, an SOC group, and phase-in periods to exclude wounds destined to heal with SOC, possibly contributing to a higher healing rate compared with pivotal trials for each agent (18).

Identifying the product category that can improve healing most effectively, with the best cost to heal, remains a challenge (19). Nonbiased studies directly comparing the efficacy of human cells, tissue, and cellular and tissue-based products to provide evidence of superiority of more costly CMs are lacking, because industries are reluctant to sponsor head-to-head trials, and other funding agencies prefer to sponsor hypothesis-driven studies over noninferiority studies. This VA Merit Award–funded study addresses this knowledge gap by comparing the efficacy of CM and ACM products relative to SOC. We hypothesized a marginal difference in efficacy between CM versus ACM compared with SOC for healing DFUs.

This randomized, single-blinded, three-arm controlled trial was conducted at the VA Northern California Health Care System (VANCHCS). Patients were recruited from six satellite clinics, and all visits were completed at the Sacramento VA Medical Center. The trial (ClinicalTrials.gov identifier NCT01450943) was performed according to the principles of the Declaration of Helsinki, the ethical principles for Good Clinical Practice, and applicable local regulations. The VANCHCS Institutional Review Board approved the study. All participants provided written informed consent before study screening and were free to opt in or out of the study whenever they wished. The full study protocol has been previously published (20–22).

Eligible participants between ages 18 and 85 years with documented diabetes and a full-thickness ulcer between 0.5 and 25 cm² present for at least 4 weeks before enrollment that did not extend to tendon, ligament, capsule, bone, or deep fascia (University of Texas grade 1A), located below the malleolus, were recruited. Additional inclusion and exclusion criteria have been previously published (21).

Participants were randomly assigned by sex, race, and age-group to one of the three treatment arms using a block stratification system generated by the biostatistician with intention-to-treat (ITT) analysis. Enrolled and randomly assigned participants received Dermagraft (CM; Organogenesis, Canton, MA), Oasis (ACM; Smith & Nephew, Andover, MA), or SOC alone. Given obvious differences in the preparation of the matrix products (CM required thawing, whereas ACM was at room temperature), blinding the clinicians was not possible; however, patients had no prior knowledge of the products and were thus single blinded.

The study schematic is provided in Fig. 1A. After a 2-week run-in period where patients whose ulcers improved or enlarged >40% with SOC were screen failed, patients who met the inclusion criteria were randomly assigned to one of three groups. All three groups received weekly SOC. For the two treatment groups, in addition to SOC, either the Dermagraft or Oasis matrix was applied to the wound after sharp debridement during weekly visits for up to eight visits until the primary end point, after which only SOC was provided until the secondary end point. The SOC protocol comprised sharp wound-bed debridement (removal of all necrotic tissue and hyperkeratotic margins until bleeding wound bed was observed). In addition, Iodosorb gel (Smith & Nephew, Andover, MA), a cadexomer bead complexed to iodine for sustained slow release, was added to SOC in all groups because of the strong high-quality clinical evidence that it improves healing in DFUs (23,24). Then, the wound was dressed with Adaptic nonadherent dressing (Johnson & Johnson, Gargrave, U.K.), cut to size, followed by gauze, which also served as secondary dressing for the two intervention arms.

For offloading, all patients enrolled in the study were initially provided with a removable walking boot or diabetic conformer Bledsoe boot (Bledsoe Brace Systems and Viscent, LLC, Grand Prairie, TX), modified to offload plantar ulcers with aperture creation in a trilaminar layer Plastazote insole, or offloaded using padding. In cases where participants did not tolerate the offloading boot, they were provided with an alternative offloading device that was tailored to their needs, which could include felt-to-foam cutout adhesives, custom-molded offloading healing shoes, custom orthotics, total contact casts, or assistive devices such as walkers, roll-about scooters, crutches, or wheelchairs. During weekly visits, investigators assessed adherence to offloading by examining the condition of dressings and devices for signs of wear and tear. Adjustments were made with appropriate offloading materials as needed. Additionally, participants were asked a weekly question about their adherence. Patients exclusively using wheelchairs were identified to differentiate fully ambulatory patients from those with limited ambulatory capacity.

Images of the wounds were captured at each visit after debridement, deidentified, and uploaded to an external server, Silhouette Central by Aranz Medical (Christchurch, New Zealand). A clinical board-certified podiatrist was the independent evaluator who had no knowledge of patients’ assignment groups and reviewed these images to determine whether the wound had closed and ensure that the margins were traced accurately. A wound was deemed healed if there was 100% re-epithelialization of the ulcer without drainage, as determined by image analysis and confirmatory clinical observation in the medical record that there was no exudate; this was again confirmed at the 2-week confirmatory visit (per protocol) (Fig. 1A). Procedures are further detailed in the published protocol and interim report (20–22).

The primary outcome was the percentage of wounds healed by 12 weeks of treatment. Secondary outcomes included the percentage of wounds healed by week 28, healing rate, cost effectiveness, association of wound healing with demographic factors and/or comorbidities, recurrence, incidence of adverse events, and quality-of-life assessments.

The original sample size was estimated based on 80% power, with a two-sided significance level of 0.05, to detect a difference in the incidence of ulcer closure between the two study groups and SOC (Stplan 4.5 statistical software; Department of Biomathematics, University of Texas MD Anderson Cancer Center, Houston, TX). Based on the reported outcomes for similar studies, we expected that 50% of the CM and ACM groups and ∼25% of the SOC group would reach complete ulcer closure by treatment week 12 (11,12). Therefore, the expected difference was 25% between the SOC and treatment groups.

At the time of the interim analysis previously reported, 11 (64.7%) of 17 had healed in the Dermagraft group, 15 (78.9%) of 19 had healed in the Oasis group, and 14 (73.7%) of 19 had healed in the SOC group (22). This suggested that in our study, the SOC group would do much better than we expected, so the original sample-size calculations should be readdressed, as was planned for in the interim analysis. In addition, the largest difference was between the Dermagraft group and the Oasis group, which suggested that further accrual should be only into those two groups and not the SOC group. This finding also suggested that the much less expensive Oasis graft might be noninferior to the Dermagraft, adding another hypothesis test to the original test comparing all three groups.

Interim analysis resulted in a change of sample size to test for noninferiority, by constructing the usual asymptotic z 95% CI for the difference of proportions healed at 12 weeks, p_Oasis − p_Dermagraft, and concluding noninferiority if the lower limit of the CI were greater than −0.20. We formalized this (in accordance with FDA guidance), showing that at the interim analysis, the rate of wound closure using Oasis was 20% better than the rate of wound closure with Dermagraft.

The primary and secondary outcomes were analyzed using a χ² test. A Kaplan-Meier survival analysis determined the mean time to complete healing. An exploratory analysis compared changes in wound size over time using an ANOVA test. All data were analyzed per ITT, with participants who withdrew either voluntarily or because of an adverse event included in the ITT analysis as not healed.

The primary statistical design and interim a priori analysis included in our original protocol were previously published (20–22). The software used for analysis was R version 3.6.1 (25). Data were analyzed by D.R. and N.T.Y. The Dermagraft Oasis Longitudinal Comparative Efficacy Study (DOLCE) was registered with ClinicalTrials.gov (identifier NCT01450943).

We will ensure that all applicable data are available to the broader scientific community per FAIR (Findable, Accessible, Interoperable, and Reusable) data principles through established repositories in the VA system.

From 1 October 2011 through 31 March 2018, ∼600 patients at VANCHCS with DFUs were screened for eligibility. After exclusion of those not meeting the inclusion/exclusion criteria, 138 patients enrolled in the study, and 117 were randomly assigned; 21 withdrew from the study, including seven who withdrew after reaching the primary end point (Fig. 1B).

Demographics and other patient characteristics are reported in Table 1. Participant ages ranged from 46 to 85 years (mean age 62 years), with 11 Black, five Native American, and 91 White participants. The mean BMI was 35.4 kg/m², mean diabetes duration was 14.5 years, and mean HbA_1c was 8.43%. Baseline ulcer data, including duration, size, and location, are reported in Table 1. All wounds were on the plantar surface of the foot. No statistically significant differences in demographic or ulcer variables were observed between groups. No significant differences in offloading adherence were reported by groups, as reported in Table 1. On average, the CM group required 6.73 matrix applications, whereas the ACM group required 6.31; this difference was not significant.

The primary (week 12) study end point was reached by 41, 48, and 28 participants in the CM, ACM, and SOC groups, respectively (N = 117), using ITT analysis (Table 2). Complete healing had occurred in 69 (59%) of the 117 patients at 12 weeks (Table 2), including 20 (49%) in the CM group and 33 (69%) in the ACM group (odds ratio [OR] 0.44; 95% CI 0.17–1.12). Sixteen participants (57%) in the SOC group had healed by 12 weeks (P = 0.16 by χ² test). Seventy (59%) had healed by 28 weeks in all groups (Table 2), including 25 (61%) in the CM group and 27 (56%) in the ACM group (OR 1.21; 95% CI 0.48–3.10). Eighteen (64%) in the SOC group had healed by 28 weeks (P = 0.78 by χ² test). We found the average wound duration and healing outcomes differed significantly (P = 0.013). As expected, participants with shorter-duration wounds were more likely to heal than those with longer-duration wounds.

The final 28-week three-group analysis revealed no statistically significant differences in wound healing percentages across groups. However, at the primary end point (12 weeks), the noninferiority analysis showed that the ACM group did achieve statistical noninferiority. We conducted a noninferiority test on these final primary end point (12 weeks) data with a margin of 10%. A result of noninferiority required showing that the percentage healed with ACM was not worse than 10 percentage points less than the healing rate of CM. We computed a two-sided 95% CI for the healing rate of ACM minus the healing rate of CM, the realized value of which was 0.6875 − 0.4878 = 0.1997. Using the R routine prop.test(), we found that the two-sided CI for the ACM healing rate minus the CM healing rate was (−0.0244 to 0.4238), showing that the ACM healing rate was at most 2.44 percentage points worse than the CM healing rate and might have been as many as 42 percentage points better. This finding implies that at 12 weeks, the ACM healing rate was at most 2.44 percentage points worse than the CM healing rate and might have been as many as 42 percentage points better. This difference did not persist to the 28-week mark, but earlier healing is still important, even if the healing eventually occurs equally later in time across all the groups.

The Kaplan-Meier survivorship analysis for time to complete healing is presented in Fig. 2. Although the ACM group seemed to heal more rapidly, no statistically significant difference in time to wound healing was calculated between groups (P = 0.71). The mean time to healing was 15.4, 12.7, and 14.1 weeks in the CM, ACM, and SOC groups, respectively (P = 0.69). Each of the treatment arms showed a statistically significant reduction in wound area from treatment week 1 to 12 and week 1 to 28 (P < 0.001). The corresponding differences were 1.54 (95% CI 0.74–2.34) and 1.65 (95% CI 0.83–2.47) for the CM group, 2.38 (95% CI 1.30–3.46) and 2.53 (95% CI 1.37–3.69) for the ACM group, and 1.08 (95% CI 0.54–1.63) and 1.33 (95% CI 0.87–1.80) for the SOC group. The mean wound sizes compared between week 1–15 (0.38 cm²) and week 28 (0.16 cm²) were similar among all three groups (week 15: 95% CI 1.25–2.30; week 19: 95% CI 1.28–2.49). When data were analyzed by categorizing based on wound size (≤1 and >1 cm²) and healing, we found no statistical differences between the groups (Supplementary Table 2). We also found no difference in healing when categorized by ankle-brachial index ≤0.9 or >0.9 (Supplementary Table 3).

A total of seven participants (CM n = 2; ACM n = 3; SOC n = 2) had recurrence of ulceration within the 2-week healing confirmation period; therefore, these patients continued in the study. Participants were only deemed to have healed if they remained healed at the 2-week confirmatory visit. During the 3-year postactive phase surveillance period, 18 (46%) of those in the CM group, 19 (40%) in the ACM group, and 10 (36%) in the SOC group had recurrent ulceration at the site of the original ulcer (P = 0.69) (Supplementary Table 1).

Serious adverse events, including any soft tissue infection or osteomyelitis, were recorded during all study phases (Supplementary Table 1). Two participants in the CM group, three in the ACM group, and one in the SOC group developed infections at the wound site, requiring hospitalization, and all recovered and went on to experience complete wound healing. There were no reports of osteomyelitis related to the ulcer site in any of the three groups. One participant in the CM group underwent a below-the-knee amputation unrelated to the study site (infection of the contralateral leg). No other foot or leg amputations were performed in any participants during the 7-month study period or the 3-year surveillance period. No participant deaths occurred during the study period, and those that occurred in the 3-year surveillance period were unrelated to the study. No allergic reactions were observed.

Although participants in both the CM and ACM groups averaged better scores on pertinent 36-Item Short-Form Survey (SF-36) questions at the end of the study relative to their baseline scores, there were no significant differences in responses to any of the SF-36 questions when comparing treatment groups from the start to the end of the study.

The cost of the CM to the VA system during the study period was $1,081.50 per unit application, and the label recommendation is for eight applications. The cost of the ACM to the VA system during the study was $107.57 per application, 10-fold lower per application than the CM. Given these prices, it would take 80 weekly applications of the ACM to reach the cost equivalence of eight CM applications.

When comparing clinical results of diabetic wound healing using SOC, a CM, or an ACM, it is important to note that the situation is not symmetric; CM products are often far more expensive than SOC or ACM products. Consequently, CM products should be routinely used only if there is sufficient evidence that they produce improved healing performance relative to other modalities. The main comparison of importance here was CM versus ACM. In both the primary and secondary outcomes, the χ² test among the three groups showed no significant differences. From the noninferiority data at 12 weeks, the direct comparison of CM with ACM demonstrated that ACM is noninferior (faster healing rate with ACM), but that difference was not apparent at 28 weeks, when all three arms had similar healing rates. After the failure of SOC, some treatment algorithms recommend initiating therapy with one of the matrix products (26–28). Because CM costs considerably more, the recommendation for its use should require evidence that it is superior to ACM.

The secondary outcome results lead to the same conclusion. The test of whether the matrices do not differ in healing rate had a P value of 0.77. The comparison of the two matrix products with the alternative hypothesis that CM is better had a P value of 0.44, and the test with the alternative hypothesis that ACM is better had a P value of 0.75. The somewhat surprising lack of superiority of CM may have been due to differences in the ECM composition. There are no data to support or refute this hypothesis, and analysis of the individual components of the respective matrices and their proreparative function in vitro would need to be carried out to do so. However, given the cost difference, this result implies that ACMs are the more cost-effective choice.

Despite human cells, tissue, and cellular and tissue-based products being increasingly marketed for the treatment of chronic wounds, few have been evaluated in RCTs, and even fewer have been compared with SOC (7,11,12,15,29–31). To our knowledge, our study is the largest head-to-head, investigator-initiated, non–industry-sponsored, randomized comparative efficacy trial comparing CM with ACM for DFU treatment. Consistent with the present findings, an earlier RCT with 26 patients found Oasis to be more cost effective than Dermagraft for DFU closure (32). Similarly, another group retrospectively reviewed the percentage of wounds healed at 90 days in 13,193 DFUs, revealing that Oasis was superior to, and less costly than, Dermagraft (33).

Unlike our findings, some RCTs comparing Dermagraft or Oasis individually with SOC have reported improvement over SOC in full-thickness ulcers (11,12,34). In those studies, however, healing with SOC was reported to be only between 18% and 32%, whereas healing in our SOC group was 57%, which is higher than in those earlier studies, perhaps because of changes in SOC that have evolved in the intervening decade (11,12,34). Other shortcomings of the aforementioned studies include small sample sizes, absence of a blinded observer, and industry sponsorship possibly promoting unintentional bias (9). Additionally, a skew in published data was noted, suggesting that studies with negative data were less likely to be published (9). Other factors such as wound size variation, duration, and grade can also affect outcomes of RCTs. We addressed the challenges of previous studies with a larger sample size, standardized wound protocol, blinded observer, intensive communication with participants to increase engagement and compliance, and absence of industry sponsorship. We also achieved higher healing rates in all groups compared with previously mentioned RCTs, including SOC. This suggests that excellent SOC may replace the need for any such product for a stalled wound. Improvement in quality of care and routine offloading with the most effective method rather than the most convenient method may be contributing to the trend of an increased percentage of healing in literature (6,8,35).

To further elucidate why our SOC group had higher rates of healing, we identified several key factors. We adhered to evidence-based protocols in treating diabetic foot ulcers (eg, sharp debridement of the wound and consistent use of dressing/offloading protocols), which provided a consistent and reliable approach across all patient encounters: the same clinical team providing consistent care; educating the patient on the wound healing/progress made each week, frequently checking on compliance with dressing changes/offloading methods, with extended time spent with the SOC group to enhance their engagement; and regularly monitoring and evaluating the wound care practices by routinely calling patients at home. We quickly recognized any deviations from established SOC and addressed them to ensure consistent high-quality patient care. Participants were provided a direct 24/7 contact phone line to our study coordinators if any concerns developed, and issues were addressed promptly. If necessary, same-day wound evaluation was performed.

Our 3-year ulcer recurrence rates are comparable to those in studies reporting 1-year recurrence rates of 25–40% (5,36,37). Although the percentage of healed wounds increased in the CM and SOC groups from week 12 to 28, it decreased in the ACM group from 69% to 56%. Three wounds recurred in the ACM group, and three patients whose wounds had previously healed withdrew from the study before the secondary end point. No participants required amputation of the study limb in 3 years of follow-up, a rate lower than predicted, because the 3-year probability for a first amputation is 13% (38,39).

By expanding recruitment to multiple sites, adjusting inclusion and exclusion criteria (21), and adjusting sample size based on a noninferiority test at interim analysis, our trial was able to achieve adequate power for statistical analysis, a common limitation in previous trials. Additionally, the centralized site for conducting the study allowed for consistent wound care. Another strength of the study was the study population that closely resembles the U.S. population with type 2 diabetes (22).

One shortcoming of our study is the inability to blind the investigators, given the nature of the matrix product. However, as noted by others, this likely does not influence end points such as ulcer healing, which was captured by the Aranz camera three-dimensional wound system and assessed by a scorer blinded to the treatment group (9). Another limitation, common to all DFU studies, is the inability to control participants’ adherence to offloading methods. Although most participants reported adherence with no significant differences found, self-reporting bias was likely, because participants are less likely to admit being noncompliant. Additionally, this study was limited to full-thickness ulcers, the label indication for the products examined, so results cannot be extrapolated to other ulcers. This study was limited to the two tested matrices, prototypes of the CM and ACM categories; other matrices brought into clinical use in the future, including newer evolving technologies incorporating placental or amniotic membranes may confer additional advantages, including limitation of sepsis and limb preservation, adverse events that were not observed in the study.

In conclusion, this investigator-initiated federally funded trial found similar clinical efficacy in healing rates between CM and ACM products in full-thickness DFUs, currently limiting the support for more costly cellular products. Moreover, rigorously administered standard weekly wound care can achieve healing rates comparable to those of CMs or ACMs, challenging the paradigm established by prior industry-sponsored studies and highlighting the potential inherent bias when SOC is used as a comparator with other products. Minimizing the use of products that do not demonstrate evidence of improvement in healing full-thickness DFUs can reduce the economic burden. Although these products play a crucial role in wound management, the importance of close follow-up should also be emphasized. Regular monitoring and assessment by wound care experts can significantly affect the healing process and outcomes of wounds, often more than the specific products used. Close follow-up allows for timely adjustments to treatment plans and ensures adherence to prescribed care regimens, which are critical for successful wound healing. Ultimately, improving SOC provides the best outcome for DFU treatment, but if selecting matrices, ACM seems to be the more cost-effective approach for healing DFUs.

Acknowledgments. The authors thank Justine Avery Arizabal for the graphical abstract. The authors acknowledge the seminal contributions of Dr. Huong L. Le to the development of this project, with the overarching goal of providing the best care to patients. The authors also thank the many patients who participated in this trial.

Funding. This work was supported by VA Merit Award CX000369 (S.D. and R.R.I.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. H.S. and R.K. researched data, contributed to the discussion, and wrote the first draft of the manuscript. C.T.-F. and H.L.-T. were involved in the collection, analysis, and interpretation of the data. K.W. was involved in data collection and contributed to the first draft of the manuscript. P.S.L., M.D., R.E.L., and D.R. researched data, contributed to the discussion, and reviewed and edited the manuscript. Biostatisticians N.T.Y., C.-S.L., and D.R. contributed to the study design/sample size calculation and data analysis. S.D. and R.R.I. designed the study and oversaw the collection, management, analysis, and interpretation of the data. All authors approved the final version of the manuscript. S.D. and R.R.I. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and accuracy of the data analysis.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Steven E. Kahn and Rodica Pop-Busui.