Topical Erythropoietin Treatment Accelerates the Healing of Cutaneous Burn Wounds in Diabetic Pigs Through an Aquaporin-3–Dependent Mechanism

According to the U.S. Centers for Disease Control and Prevention, >400 million people worldwide have diabetes. The Centers for Disease Control and Prevention also estimates that 5% of these individuals will develop a diabetic skin ulcer (DSU) and that 1% will require a lower-extremity amputation. Despite many advances in wound care and management (1), wound healing in diabetes is delayed because all phases of the orchestrated cascade of cellular and biochemical events of wound healing are disrupted (2–4). Additionally, the healing of a DSU is delayed because of impaired angiogenesis, insufficient blood flow, increased inflammation, diminished proliferation of fibroblasts (5), and reduced reepithelialization by keratinocytes (6–8).

The glycoprotein hormone erythropoietin (EPO) regulates red blood cell mass and is an approved drug for treating anemia. EPO also has nonhematopoietic targets in the skin, and we have shown (9) that these targets participate in the healing of skin wounds. We previously reported that the healing of cutaneous wounds in rats and mice with experimentally induced diabetes is accelerated after the topical application of recombinant human EPO to the cutaneous wounds by stimulating angiogenesis, reepithelialization, and collagen deposition, and by suppressing the inflammatory response and apoptosis (10). The beneficial actions of EPO on wound healing in diabetes are complemented by fibronectin (FN). FN facilitates the formation of the provisional wound matrix and prevents its dissociation (11).

Aquaporins (AQPs) are integral membrane proteins whose function is to regulate intracellular fluid hemostasis by enabling the transport of water and glycerol. AQPs are expressed in the plasma membranes of keratinocytes in the basal layer of the skin and the medullary collecting ducts of the kidney (12). Downregulated expression of AQPs may be the cause of the reduction in urinary-concentrating ability in individuals with acute renal failure, and EPO can prevent this downregulation (13). AQP3 is the AQP that is expressed in the skin (14), where it facilitates cell migration and proliferation and reepithelialization during wound healing (15–18). A positive role for moisture in healing skin wounds was first shown in 1962, when Winter (19) investigated scab formation and the rate of epithelialization of superficial wounds in pig skin and reported that moist wounds heal faster than dry ones. As a corollary, dryness of the skin of the feet correlates with foot ulceration in patients with diabetes (20). A specific role for AQP3 in diabetic wound healing was posited when it was found that AQP3 is downregulated in the regenerating epidermis during the healing of full-thickness cutaneous wounds in rats with diabetes (21). The work of these investigators and our results indicated that EPO could be used to stimulate the healing of nonhealing wounds. Such findings also suggest the existence of a causal relationship between impaired AQP3 expression and delayed reepithelialization in diabetes, which involves impaired movement and proliferation of those cells that participate in angiogenesis, reduced production of the extracellular matrix (ECM) by fibroblasts, and failure of keratinocytes to reepithelialize a cutaneous skin wound.

It is against this background that we posited that the therapeutically beneficial action of EPO on the healing of a DSU is due in part to the ability of EPO to stimulate AQP3 expression in skin. We tested this hypothesis in pigs with experimentally induced type 1 diabetes and a partial thickness skin burn.

The study comprised four 60-kg healthy female pigs (Sus domesticus) purchased from Lahav Institute (Kibbutz Lahav, Israel). The pigs were singly housed in pens in a room with an artificial 12-h light/dark cycle and had free access to a standard laboratory chow and water. All procedures were performed in the Technion according to the Israeli national legislation on the use of animals for experimental purposes.

Diabetes induction, creation of the burn wounds, wound treatment, dressing changes, and data collection were performed under general anesthesia. An intravenous catheter was permanently placed in the right jugular vein in the four pigs for blood sampling during the investigation. Heart rate, blood oxygen saturation, and the body temperature of the pig were monitored. On day 14, the last day of the experiment, specimens were collected and the pigs were then humanely killed.

Diabetes was induced in two pigs using streptozotocin (200 mg/kg; Alexis Biochemicals) according to a previously described protocol (22). Blood glucose levels were checked every 15 min for 6 h after streptozotocin administration and at least twice daily during the study period using a glucometer (FreeStyle FREEDOM Lite; Abbott). Long-acting insulin (24 international units [IU] Lantus; Aventis Pharmaceuticals) was injected intravenously to maintain the fasting blood glucose levels of the diabetic pigs between 300 and 400 mg/dL.

Diabetes was maintained for 1 month before partial thickness skin burn wounds were created. The bristles of the dorsal skin on each side of the vertebral column of each pig were removed using VEET depilatory cream (Reckitt Benckiser) before creating 12-cm² circular partial thickness skin burn wounds using an aseptic technique. The burn wounds were created using a previously described method (23). Briefly, four cylindrical brass rods, each with an ∼3.8-cm diameter and weighing 358 g, were placed in hot water (92°C) for 2 min. Four groups of six partial thickness burn wounds were created by placing of the rod perpendicular on the dorsal skin of each pig for 20 s with no additional pressure “producing partial thickness third-degree burns.” In the two diabetic pigs, an additional group of six partial thickness burn wounds was created.

Six different gels for topical wound treatment were prepared in the Remedor Biomed Ltd. laboratory in accordance with the following recommendations by the U.S. Pharmacopeial Convention: 1) a gel that contained no active ingredients (vehicle gel); 2) a gel that contained 2,000 IU/g EPO (high-dose EPO); 3) a gel that contained 500 IU/g EPO (low-dose EPO); 4) a gel that contained 300 μg/g FN; 5) a gel that contained 2,000 IU/g EPO and 300 μg/g FN; and 6) a gel that contained 0.1 mmol/L mercuric chloride (HgCl₂) and 2,000 IU/g EPO. Recombinant human EPO was purchased as an injection (EPREX 40,000 IU; Janssen-Cilag Ltd., High Wycombe, U.K.). FN was purchased as a 1 mg/mL solution (EMD Millipore). HgCl₂ (0.1 mmol/L) was incorporated into the gel that contained 2,000 IU/g EPO in order to block cutaneous AQP3 activity in the wounds (negative control) and was purchased from Sigma-Aldrich. The results of the stability testing of the gels established that EPO and FN are stable in the gel for at least 6 months at 4°C, as determined by ELISA.

Each group of the six partial thickness burn wounds was randomly assigned to be treated with one of the following topical gels: the vehicle-containing gel; the high-dose EPO-containing gel; the FN-containing gel; and the EPO/FN-containing gel. The additional group of the burn wounds in one diabetic pig was treated with the EPO/HgCl₂-containing gel, and the additional group of burn wounds in the second diabetic pig was treated with a low-dose EPO-containing gel (low-dose EPO). A simple randomization sequence was generated by computer software. After allocation of the six partial thickness burn wounds in four different groups in each pig, gel (3 g) was topically applied to each wound every 2 days of the 14-day study period. In order to prevent removal of the gel after its application by rubbing and to protect the burn wounds between treatments, the treated wounds were covered with nonadherent gauze pads, which were stabilized by Tensoplast Elastic Adhesive Bandaging (Smith & Nephew).

On each treatment day, pigs were weighed; each wound was photographed using a 12-megapixel digital camera (Olympus); and a blood sample was collected for determining the red blood cell, leukocyte, and platelet counts, and the plasma hemoglobin and glycated hemoglobin (HbA_1c) levels. Punch biopsy specimens from randomly selected areas of the regenerating skin of the vehicle-treated and treated burn wounds and the uninjured skin of the pigs were collected on days 2, 7, and 14 using a 6-mm circular blade. Samples of each biopsy specimen were fixed immediately in 10% neutral buffered formalin for histological determination of the microvascular density (MVD) or stored in liquid nitrogen for immunohistochemistry, Western blot analysis, ELISA, and PCR. The slides were examined under a Nikon Eclipse E800 Upright Microscope, and images of the sections were captured and analyzed by Metamorph Image Analysis Software (Nikon Instruments Inc.).

The wound closure rate was calculated from the area of reepithelialized tissue in the burn wound. In order to calculate the rate, the area of each burn wound was categorized into the following three areas: 1) a scab area where the burnt skin becomes a rigid crust after creation of the burn injury; 2) a red area where the scab can be detached but has no epithelial cells; and 3) a white area where the scab can be detached and contains new epithelial cells. The wound closure rate was calculated by measuring the white area on days 2, 4, 7, 9, 11, and 14. To this end, transparent paper was placed over each wound, and the shape of the white area was drawn on the paper. The transparent paper was then superimposed onto a 1-mm² graph paper in order to measure the white area in the wound. The rate of time-dependent changes in the size of the white area were calculated using the following formula:

Blood flow in the wounds was measured noninvasively by a laser Doppler perfusion imaging system (PeriScan PIM 2 System; Perimed) during the study period according to the manufacturer instructions. Perfusion in a region of interest (ROI) is measured on a scale of six colors in which dark blue depicts the lowest perfusion rate and red depicts the highest perfusion rate using PIMSoft software for blood perfusion imaging (Lisca Development AB). For each burn wound, the ROI was a 12-cm² circle that was drawn around the initial burn wound, and the blood flow in the wound is the average value of all colors in the ROI.

The MVD and the extent of angiogenesis in the regenerating skin of the healing wounds was determined by staining 5-µm thick sections of the formalin-maintained samples punch biopsies of the wounds with CD31 (R&D Systems). The number of capillaries in the regenerating skin in each wound site was counted in five random microscopic fields (×20 magnification).

AQP3 was detected in wound-free healthy and diabetic pig skin 1 month after diabetes induction and 1 day before creation of the burn wounds and in the regenerating skin of the wounds during the study period by immunohistochemistry and immunofluorescence. For AQP3 immunohistochemistry, 5-µm-thick sections were stained with rabbit polyclonal AQP3 primary antibody (Santa Cruz Biotechnology). For confirming the presence of AQP3 by immunofluorescence, another set of the AQP3-stained sections were counterstained with the nuclear stain TOPRO-3 (Invitrogen) and examined under a confocal microscope (BIO-RAD).

The amount of collagen in the regenerating skin was determined in specimens that were stained with Masson’s trichrome (Sigma-Aldrich). Using this method, collagen fibers stained blue, nuclei stained black, and cytoplasm and muscle fibers stained red.

Hydroxyproline (HP), an amino acid constituent of type I collagen, was used as a marker and an indicator of the amount of collagen in the skin specimens from the regenerating skin of the burn wounds. The amount of HP was determined using a previously described protocol (10).

Since hyaluronic acid (HA) is linked to skin strength and hydration, the HA amount in the regenerating skin of the burn wound tissues was determined in specimens that were collected from the nondiabetic and diabetic pigs using the Hyaluronan Quantikine ELISA Kit (R&D Systems) according to the manufacturer protocol.

The expression levels of AQP3, endothelial nitric oxide synthase (eNOS), HA synthase (HAS) 1, and HAS2 were determined in the regenerating skin of the burn wound tissues. Briefly, tissue extracts were prepared using radioimmunoprecipitation assay buffer (Elpis-Biotech). Proteins from tissue lysates were separated by 10% SDS-PAGE and then transferred to nitrocellulose membranes. The membranes were probed with indicated antibodies. Protein expression levels were detected by densitometry using the Immun-Star horseradish peroxidase chemiluminescence detection system (BIO-RAD). The results were expressed as a percentage of the average protein level of duplicate readings in the vehicle-treated burn wounds of the healthy pigs (100%).

Total RNA was extracted from homogenates prepared from regenerating skin of the burn wound tissues using the MasterPure RNA Purification Kit (EPICENTRE Biotechnologies). cDNA was generated by absolute qPCR mixes reverse transcription reagents (ABgene) and analyzed by qPCR using SYBR Green PCR Master Mix (Molecular Probes) in a Rotor-Gene 6000 Cycler (Corbett Life Science). The results were expressed as a percentage change from control after normalization to an endogenous reference gene (GAPDH).

All statistical analyses were performed using a computerized statistical program (GraphPad Prism version 5.0; GraphPad Software Inc.), and all data are presented as the mean or percentage ± SD. Statistical significance was set at 5%. A two-tailed Student t test was used to compare study parameters of the healthy and diabetic pigs, and a two-way ANOVA with Bonferroni correction to control for type I error was used for multiple comparisons. The wound closure rates among groups were compared and analyzed using a one-way ANOVA with Tukey post hoc test and a two-way ANOVA with Bonferroni correction to control for type I error in the multiple comparisons.

Topical treatment of the burns with the vehicle-, FN-, EPO- and EPO/FN-containing gels did not change 1) the red blood cell, leukocyte, or platelet counts and the elevated blood glucose levels in the diabetic pigs and 2) the blood hemoglobin and HbA_1c levels in the control and diabetic pigs (Table 1).

The wound closure and blood flow rates of the vehicle-treated diabetic wounds were significantly lower than those of the vehicle-treated healthy wounds from day 2 onward. The wound closure and blood flow rates of the EPO-treated, the FN-treated, and the EPO/FN-treated diabetic wounds were significantly greater than those of the vehicle-treated diabetic wounds. The most significant differences in wound closure rates were detected between the vehicle-treated and the EPO/FN-treated diabetic wounds. From day 4 onward, the wound closure rate of the EPO/FN-treated wounds was significantly faster than those of the FN-treated and the EPO-treated diabetic wounds. The blood flow in the EPO-treated diabetic wounds was not significantly different from that of the EPO/FN-treated diabetic wounds, and these two blood flows were significantly higher than those in the vehicle-treated and FN-treated diabetic wounds (Fig. 1A–C).

The wound closure and blood flow rates of the low-dose EPO-treated diabetic wounds were significantly greater than those of the vehicle-treated diabetic wounds, and the wound closure and blood flow rates of the high-dose EPO-treated diabetic wounds were significantly greater than those of the low-dose EPO-treated diabetic wounds from day 9 onward (Fig. 1D–F).

The wound closure and blood flow rates of the vehicle-treated and FN-treated wounds of healthy pigs were very similar. From day 4 onward, the wound closure and blood flow rates of the EPO-treated nondiabetic wounds were significantly greater than those of the vehicle-treated and FN-treated nondiabetic wounds. The most significant differences in wound closure and blood flow rates were detected between the EPO/FN-treated wounds and the vehicle-treated and the FN-treated wounds in the healthy pigs over the period from day 4 to day 11 (Fig. 1G–I).

The MVD and eNOS expression levels in the vehicle-treated diabetic wounds were lower than those in the vehicle-treated nondiabetic wounds after 14 days. Topical FN treatment had no effect on the MVD and eNOS expression levels in the diabetic and nondiabetic wounds. On the other hand, topical EPO treatment for 14 days significantly increases the MVD and eNOS expression levels in the diabetic and nondiabetic wounds (Fig. 2A–E). In the EPO/FN-treated nondiabetic wounds, the MVD and eNOS expression levels were not significantly different from those in the EPO-treated nondiabetic wounds (Fig. 2B and C). In contrast, the MVD and eNOS expression levels of the EPO/FN-treated diabetic wounds were significantly higher than those in the EPO-treated diabetic wounds (Fig. 2D and E).

The amounts of HP and HA and the expression levels of the two HASs, HAS1 and HAS2, in the vehicle-treated diabetic wounds were lower than those in the vehicle-treated nondiabetic wounds. Topical FN treatment for 14 days did not change the HP and HA amounts and the HAS1 and HAS2 expression levels in the diabetic and nondiabetic wounds. On the other hand, topical EPO treatment significantly increased the HP and HA amounts and the HAS1 and HAS2 expression levels in diabetic and nondiabetic wounds (Fig. 3A–G). HP amounts in the EPO/FN-treated nondiabetic and diabetic wounds were significantly higher than those in the EPO-treated nondiabetic and diabetic wounds, respectively (Fig. 3B and C). The HA amount and the HAS1 and HAS2 expression levels in the EPO/FN-treated nondiabetic wounds were similar to those of the EPO-treated nondiabetic wounds (Fig. 3D and F). In contrast, the HA amount and the HAS1 and HAS2 expression levels in the EPO/FN-treated diabetic wounds were significantly higher than those in the EPO-treated diabetic wounds (Fig. 3E and G).

AQP3 protein and mRNA expression levels in wound-free diabetic pig skin were significantly lower (P < 0.01 for both) than those in wound-free healthy pig skin (Fig. 4A–G).

After 14 days, AQP3 protein and mRNA expression levels in the vehicle-treated diabetic wounds were significantly lower than those in the vehicle-treated nondiabetic wounds. In the diabetic and control pigs, AQP3 protein and mRNA expression 1) in the EPO-treated wounds were significantly higher than those of the vehicle-treated and FN-treated wounds, 2) in the EPO/FN-treated wounds were significantly higher than that of the EPO-treated wounds, and 3) in the vehicle-treated and FN-treated wounds were not different from each other (Fig. 4J–O).

We used the Pearson correlation to investigate the relationships among AQP3 protein expression levels, the extent of angiogenesis, and the amounts of HP and HA in the burn wounds of healthy and diabetic pigs. In the diabetic and healthy pigs, AQP3 protein expression levels correlated positively with the extent of angiogenesis (Fig. 5A and B), the HP amount (Fig. 5C and D), and the HA amount (Fig. 5E and F). Interestingly, these correlations were significantly stronger in the regenerating skin of the burn wounds of the diabetic pigs than those found in the regenerating skin of the burn wounds of the healthy pigs.

When AQP3 activity was blocked by HgCl₂, the wound closure rates of EPO/HgCl₂-treated diabetic wounds were significantly reduced. This reduction in wound closure rate was accompanied by a reduced blood flow (Fig. 6A and B), low eNOS expression levels, low HAS1 and HAS2 expression levels (Fig. 6C), a low extent of angiogenesis, a reduced HP amount, and a reduced HA amount. Moreover, the wound closure rates; the expression levels of eNOS, HAS1, and HAS2; and the amounts of HP and HA in the EPO/HgCl₂-treated diabetic wounds were significantly lower than those in the EPO-treated diabetic wounds (Fig. 6D–F).

The healing of a DSU is delayed because of impaired angiogenesis, reduced cutaneous cellular activity, and increased inflammatory response. In this report, we describe a new mechanism by which topical EPO accelerates the healing of a diabetic skin wound. We found that topical EPO treatment of the burns in the diabetic pigs accelerated their healing through an AQP3-dependent mechanism by stimulating angiogenesis and ECM production.

Angiogenesis, the synthesis of ECM constituents such as collagen and HA, and proper cell hydration in the wound bed are indispensable for normal wound healing. EPO stimulates endothelial cell proliferation and the secretion of angiogenic cytokines and growth factors, such as vascular endothelial growth factor, fibroblast growth factor, and IGF-I from endothelial cells and keratinocytes, to cause the sprouting of new blood vessels into the wound bed (24). In this investigation, we found that topical EPO treatment of a wound substantially increases blood flow in the regenerating skin of diabetic burn wounds as measured by laser Doppler scanning. We confirmed this effect when we measured the MVD and eNOS expression levels in the regenerating skin of diabetic burn wounds. Topical EPO treatment also resulted in significantly increased amounts of HP and HA in the diabetic burn wounds. Vedrenne et al. (25) reported the existence of a close relationship between the ECM and the synthesis of molecules that regulate attachment between cells and the ECM, angiogenesis, skin wound healing, and turnover of resident dermal fibroblasts. Collagen and HA have many functions in the ECM, one of which is to be a tissue scaffold for maintaining cellular shape and differentiation, supporting cellular movement and migration, and enabling the ECM in the dermal layer to resist compression.

AQP3 is abundant in native and reconstructed human skin epidermis, where it is primarily localized to keratinocyte plasma membranes in the epidermis and is consistent with water distribution in this tissue. In addition, water transport and water permeability studies demonstrated that AQP3 is functional in human epidermis and confers a high water permeability to viable layers of the epidermis. These results suggested that AQP3 can play a significant role in the hydration of the epidermis by preventing the formation of an osmotic gradient across viable layers of this tissue (14,26). Growing evidence demonstrated that AQP3 facilitates cell migration and proliferation and reepithelialization during wound healing (17,18), and skin restoration after an injury is boosted when AQP3 is stimulated (27). Cell hydration and a moist environment are critical for facilitating fibroblast turnover, angiogenesis, and reepithelialization by keratinocytes during wound healing. Therefore, any factor that prevents or limits local AQP3 protein expression and/or activation probably reduces the level of cell hydration and impair wound healing. It was demonstrated that AQP3 is downregulated in the regenerating epidermis during the healing of full-thickness cutaneous wounds in rats with diabetes (21). Here, we found that AQP3 expression levels were reduced in the wound-free pigs and in the regenerating skin of the diabetic pigs compared with those in the healthy pigs.

In this study, we found that the slow wound closure rate of the diabetic wounds is associated with reduced angiogenesis and low HP and HA amounts in the wound bed. HA is a very hydrophilic molecule, and this property enables it to regulate tissue hydration because it attracts and binds water. We also found that topical EPO treatment significantly increased angiogenesis and the amounts of HP and HA in the diabetic wounds and that these increases were correlated with a significant increase in AQP3 expression levels. Furthermore, we found that these correlations were stronger in the EPO-treated and EPO/FN-treated burn wounds of the diabetic pigs than those in the EPO-treated and EPO/FN-treated burn wounds of the healthy pigs. Expectedly, the inhibition of AQP3 by HgCl₂ in the burn wounds of diabetic pigs antagonized the positive actions of EPO, and this result implies that EPO-mediated stimulation of AQP3 can stimulate wound healing in diabetes. Altogether, suggest that EPO-induced acceleration of healing is mediated through AQP3-dependent mechanisms. When these mechanisms are activated, the key events of the wound-healing process, namely angiogenesis, collagen and HA synthesis, and reepithelialization, are stimulated, and the wound closure rate is accelerated.

Another interesting finding in our study was that AQP3 expression levels in the EPO/FN-treated diabetic wounds were four times higher than those in the EPO-treated diabetic wounds. This substantial increase in AQP3 expression levels in the EPO/FN-treated diabetic wounds is associated with rapid wound closure rates and elevated blood flow and a twofold increase in the MVD and the amounts of HP and HA in the wound tissues. FN has an essential function in the formation of granulation tissue during the proliferative phase of wound healing (7,15). In diabetes, a deficiency in FN and/or its degradation by proteases leads to disintegration of the provisional matrix, and reepithelialization does not occur or is delayed (11). We found that exogenous FN alone has no effect on wound closure rate and does not potentiate the accelerating action of EPO on the healing of burn wounds of healthy pigs because endogenous FN is present in normal levels and is not degraded in healthy pigs. Since endogenous FN is degraded in a DSU, we found that incorporating FN into an EPO-containing gel is desirable because FN potentiates the salutary actions of an EPO-containing gel.

The results of this study provide evidence that supports our hypothesis that topical EPO treatment of burn injuries in the skin of diabetic pigs accelerated their healing through an AQP3-dependent mechanism by stimulating angiogenesis and ECM production. The results of this study can also expect that stimulating AQP3 expression in a nonhealing ulcer by EPO further accelerates healing, perhaps by increasing cell hydration and raising the moisture levels of the wound. Increasing cell hydration and raising moisture levels facilitate interactions among the various cell types and ECM components. However, the effect of EPO on the hydration of regenerating skin cells is still under investigation and has to be illustrated. Such interactions result in proper cellular movement, migration, and differentiation and, ultimately, in the restoration of intact skin. These new and exciting findings in a diabetic pig present the possibility that the topical application of EPO may be therapeutically beneficial for stimulating the healing of a DSU by raising cutaneous AQP3 expression levels. Additional studies are now required to validate these findings, and clinical trials are needed to evaluate the safety and efficacy of topical EPO treatment in patients with diabetes and a DSU.

Acknowledgments. The authors thank Dr. Arieh Bomzon, ConsulWrite (www.consulwrite.com), for his editorial assistance in preparing the manuscript.

Funding. This study was supported by Remedor Biomed Ltd. and the Office of the Chief Scientist of Israel’s Ministry of Economy.

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

Author Contributions. S.H. designed, supervised, and carried out the experiments; performed computational analysis; analyzed all of the data; and wrote the manuscript. Y.U. designed, supervised, and carried out the experiments and analyzed all of the data. D.E. designed and carried out the experiments and analyzed all of the data. A.K. and E.D. carried out the experiments, performed computational analysis, and supervised experiments with imaging of pig tissues, as well as some immunological measurements. O.A. carried out the experiments and supervised experiments with imaging of pig tissues, as well as some immunological measurements. H.K. and M.A. prepared vehicle-, EPO-, EPO/HgCl₂-, FN-, and EPO/FN-containing gels. M.B. performed the computational analysis. R.S. and A.Z. helped to carry out the experiments. M.S. helped to carry out the experiments and supervised the experiments with laser Doppler. L.T. designed the experiments, supervised all experiments, and analyzed all of the data. P.Y.L. supervised all experiments and wrote the manuscript. S.H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.