Abnormal joint and bone wound healing in hemophilia mice is improved by extending factor IX activity after hemarthrosis

In the clinical care of hemophilia, the most common sites of bleeding are musculoskeletal. Recurrent joint bleeding leads to hemophilic arthropathy, which is the major morbidity of hemophilia. The pathogenesis of bleeding-induced joint degeneration in the past has been studied by exposing the synovium and cartilage of hemostatically normal animals to blood or blood components, either in vivo or ex vivo (eg, tissue explants).^1-3  However, it is possible that studying healing of joint wounds in animals with intact hemostatic responses does not fully recapitulate healing processes in animals with impaired hemostatic potential.^4-6  For this reason, we have recently modeled end points related to hemophilic intra-articular (IA) bleeding in hemophilia A and B mouse models. Specifically, a model of hemophilic gross joint hemorrhage induced by the puncture of the knee joint capsule with a fine gauge needle has been used by multiple research groups.^7-15  This injury model is useful because hemostatically normal mice display minimal or no pathologic changes after a single puncture of the joint capsule, but hemophilic mice develop pathologic joint changes including synovial proliferation, exuberant neoangiogenic invasion, and blood and hemosiderin iron staining. These changes mirror human bleeding-induced arthropathy.^7,8 

During the clinical management of acute hemophilic joint hemorrhage, clotting factor is typically replaced IV until there is evidence that ongoing bleeding has stopped (eg, decreasing pain, stabilized joint swelling, improving joint range of motion). There is variance in clinical practice how long factor replacement is continued, if at all, after initial hemostasis is achieved.^16,17 

The current study explored the hypothesis that abnormalities will be observed in the time course and histopathology of healing in the blood-exposed joint in hemophilia B mice (when compared with mice with normal hemostasis). We hypothesized in addition that treating the hemarthrosis with a dose of clotting factor that corrects hemostasis will not restore normal joint wound healing, which will require instead prolonged restoration of thrombin generating potential.

Nonacog β pegol (N9-GP) is a recombinant factor IX (rFIX) protein that has been modified by attaching a single 40-kDa polyethylene glycol (PEG) molecule in an effort to prolong the circulating half-life of FIX zymogen. The half-life prolongation of N9-GP has been demonstrated in human prelicensure trials and has been associated with clinical resolution of target joint bleeding when infused as once-weekly prophylaxis.^18-21  In addition, the pharmacokinetic profile of N9-GP in FIX^−/− mice has been well characterized. N9-GP has a circulating half-life of 41 hours in this strain, compared with a 17-hour half-life for unmodified rFIX (Benefix).²²  We explored the hypothesis that the use of a FIX protein that provides extended thrombin-generating potential (N9-GP) would achieve improved synovial and osteochondral wound healing in hemophilic mice when compared with treatment with unmodified rFIX.

Wild-type (WT) C57Bl/6J mice were purchased from the Jackson Laboratory (Bar Harbor, ME). FIX knockout C57Bl/6J (FIX ^−/−) male mice²³  were bred in-house, were backcrossed to Jackson Laboratory C57Bl/6J for 12 generations (12N), and were age 7 to 10 weeks at the time of investigations. All investigations were approved by the University of North Carolina, Chapel Hill Institutional Animal Care and Use Committee. Mice were anesthetized using isoflurane/O₂ for all procedures. All blood samples were collected from the retro-orbital plexus into 1:9 parts 3.2% citrated sodium and the platelet-poor plasma was stored at −80°C. The knee joint IA bleeding challenge results from the puncture of the joint capsule using a Hamilton syringe with a 30.5-G needle via a small (∼0.5 mm) incision of the skin overlying the patella as described.^7-10  Following injury, all of the mice had access to Tylenol gel for pain relief. Knee joints were collected by sectioning the femur and tibia/fibula 1 cm from the joint. Joints were fixed and decalcified using routine histologic procedures.

Three groups of mice were studied to establish the time course of histopathologic changes that occur in joint tissues over an 8-week period following articular exposure to blood in the joint puncture bleeding challenge. All mice received puncture injury of the left knee at day 0 to induce bleeding. Five microliters of normal saline (NS) was introduced into the joint space via the puncturing needle in groups of FIX^−/− mice and WT C57Bl/6J mice (“WT NS” group). Citrated whole blood was collected from a third group of hemostatically normal C57Bl/6J mice, and each animal received 5 μL of its own fresh autologous blood injected into the joint at the same time that the wounding needle was introduced (“WT + Blood” group). Animals received no replacement coagulation factor. Separate groups of mice were euthanized for histopathologic examination on day 0 (unwounded) and on days 1, 3, 5, 7, 10, 14, 28, or 56 for the wound healing time course studies (4-8 mice per group per time point) or on day 14 in the treatment studies (7-8 mice per group).

Unmodified rFIX was BeneFIX (Pfizer, Philadelphia, PA). A dose-finding study was performed to determine the most informative dose of N9-GP (supplied by Novo Nordisk A/S, Maaloev, Denmark) for demonstrating differences (should they exist) in treatments to support wound healing. Details of this dose-finding study are included in the supplemental Methods and Results and supplemental Table 1 (available on the Blood Web site). In brief, a dose of N9-GP of ∼250 U/kg protected against severe synovitis in a majority of hemophilia B mice, but did not completely protect from mild synovial inflammation and other histopathologic abnormalities. Therefore, it was expected that this dose would allow quantifiable observation of a range of therapeutic responses to N9-GP. Additionally, the detailed comparative circulating pharmacokinetics of plasma FIX antigen and activity following dosing of N9-GP and rFIX in this strain of FIX^−/− mice at this dose have been generated previously.²² 

Comparative evaluation of FIX levels in plasma and synovial fluid was performed following administration of rFIX and N9-GP as per methods previously described⁸  and detailed in the supplemental Methods.

Joint hemorrhage was induced in FIX^−/− mice or WT mice. At 20 minutes after wounding, the FIX^−/− mice were treated IV with either NS, with 250 IU/kg of N9-GP, or with 250 IU/kg of rFIX. WT mice received NS IV. Additional groups of mice received unmodified rFIX in repeated dose schedules after wounding. One group received 250 IU/kg of rFIX at the time of injury, on the following day, and again on the third day, equivalent to the “enhanced episodic” factor replacement schedule used in the Joint Outcome Study of Manco-Johnson et al.¹⁶  Another group of mice received a total of 8 replacement doses of rFIX given on an alternate-day schedule throughout 2 weeks following hemarthrosis. At 14 days after the hemarthrosis, mice were euthanized and tissues collected for histologic processing. Plasma FIX inhibitor testing was performed using the Bethesda inhibitor assay modified to detect human FIX in mouse plasma as previously described (lower level of sensitivity 1 Bethesda inhibitor unit per mL).²⁴ 

Sagittal sections of knee joints were prepared and stained with hematoxylin and eosin for synovitis grading. Hemophilic synovitis was graded in terms of synovial hyperplasia (0-3 points), vascularity (0-3 points), and the presence of discoloration, blood, villi, or cartilage erosion (0 or 1 point for each), resulting in a combined score of 0 to 10 points for increasing pathology, according to the system validated by Valentino and Hakobyan,²⁵  as previously described.^8,9  Additional sections were prepared with Prussian Blue stain for tissue iron and for immunohistochemical staining and grading for macrophages and vessels were counted by staining for endothelial cells by von Willebrand factor (VWF) as previously described by Hoffman et al⁶  with some modifications, as detailed in the supplemental Methods. For the time-course study only, joint cartilage pathology from Safranin-O–stained sections was graded to generate the Global Score of bleeding-induced joint pathology, as previously described by Narkbunnam et al⁹  (supplemental Figure 1). In light of the relatively small contribution of gross cartilage pathology to the Global Score, cartilage was additionally evaluated throughout the time course for total number of chondrocytes per high-powered field and for number of chondrocytes undergoing apoptosis per high-powered field.

Mice in the FIX treatment studies were euthanized on day 14 following hemarthrosis, and hind limbs were collected and fixed in formalin. Hind limbs were imaged using microCT at 10-µm resolution (µCT80; Scanco Medical AG, Brüttisellen, Switzerland). Analysis was performed on the trabecular bone of the injured limb inferior to the growth plate of the proximal tibia. Scanco analysis software was used to perform the histomorphometric analysis of trabecular morphology and bone mineral density (BMD), as previously described,²⁶  and to derive a quantitative expression of the degree of disruption of the articular surface of the joint.

The evolution of pathologic changes was examined by 2-way analysis of variance with Tukey’s adjustment for multiple comparisons (3 pairwise comparisons on each day). The effect of extended rFIX activity was examined by 1-way analysis of variance with Sidak’s adjustment for multiple comparisons (10 pairwise comparisons). An adjusted P value <.05 was considered a statistically significant difference. The analyses used version 6.00 for Windows (GraphPad Software, La Jolla, CA). The data are shown as the mean ± standard error of the mean.

Joint bleeding was induced by needle puncture of the left knee joint capsule, as previously described,^7-15  in cohorts of FIX^−/− and hemostatically normal WT mice. To establish that mice of both hemostatic phenotypes did experience bleeding into the IA space following this challenge, groups of mice were euthanized at 2 hours after joint puncture, and the gross tissues were examined visually and the histology examined microscopically. Blood was observed in joints of both strains, although the amount of bleeding into the FIX^−/− joints was greater than into the WT mouse joints (data not shown). Therefore, a third treatment group was added, composed of WT mice that experienced a puncture of the left knee joint capsule in an identical fashion to the other mice, but at the time of the needle puncture, 5 μL of freshly collected citrated autologous blood was injected into the joint to insure significant blood exposure (“WT + Blood” study group).

An 8-week time course of changes in the synovium and cartilage following hemarthrosis demonstrated different outcomes between hemophilic and hemostatically normal mice (supplemental Figure 1; Table 1). As early as 24 hours after injury, the clearance of blood from the synovial space in the WT and WT + Blood groups was nearly complete. This finding contrasts with the frank blood that dwelled in the joints of FIX^−/− mice through at least day 3, by which time synovial pathology was evident (Figure 1A-B; supplemental Figure 2). Any minimal changes in synovial histology in both groups of mice with normal thrombin-generating potential were not evident after 1 week, whereas pathology did not resolve in hemophilia B mice throughout the 8 weeks of observation (Table 1; Figure 1A-B; supplemental Figure 2).

Acute cartilage insult was demonstrated by the finding of apoptosis within chondrocytes of the hemophilic mouse cartilage using TUNEL stain. Chondrocyte apoptosis peaked at 1 week after knee injury, by which time a loss of the total number of chondrocytes was observed. Resolution of apoptotic chondrocytes was not observed until after week 4 in the injured FIX^−/− mice, so that chondrocyte counts did not recover by the end of 8 weeks of observation (Figure 1C-D). Neither group of injured hemostatically normal mice demonstrated chondrocyte changes.

We performed a qualitative analysis of hemophilic joint wound healing following hemarthrosis using the same scales used by Hoffman et al^6,27,28  (Table 1). Synovial iron deposition in FIX^−/− mice was obvious beginning day 7 peaking at day 28 but never clearing throughout the 8 weeks of observation. There was no iron deposition in hemostatically normal mice at any time point, with iron from the injected autologous blood effectively cleared (Figure 2A). Macrophage/monocyte residence in the joint space peaked at days 3 to 7 in both groups of hemostatically normal mice and returned to baseline after the first week. The peak inflammatory cell response was delayed and ultimately more profound in hemophilic mice (Figure 2B). Neoangiogenesis was present in some individual mice of both the WT and hemophilic mouse groups (Figure 2C). Although the quantity of neovessels in the blood-exposed injured WT mouse joint was scant and not statistically significant, VWF-stained vessels were both more prominent and more persistent in the hemophilic mice.

To establish the relative importance of initial and extended hemostatic support for healing following hemophilic joint bleeding, FIX replacement using unmodified rFIX was compared with FIX replacement using N9-GP. A dose-finding study of the effect of N9-GP given early in response to induced joint hemorrhage determined that comparing doses of 250 IU/kg of N9-GP and rFIX in subsequent studies would yield an informative range of therapeutic responses (supplemental Methods and Results and supplemental Table 1). Additionally, we examined whether monoPEGylation of FIX affected the accessibility of the protein to the intraarticular space. At 2 hours after IV administration of 250 IU/kg of either rFIX or N9-GP, the levels of FIX levels measured in the synovial fluid were similar (detailed in Table 2 and supplemental Methods and Results).

Joint hemorrhage was induced in FIX^−/− or WT mice. At the time of euthanization, FIX inhibitors (≥1 Bethesda unit) were not present in plasma of any treated mice. None of the therapies given to hemophilic mice to interrupt ongoing hemorrhage (administered 20 minutes after wounding) resulted in joint health equivalent to that seen in the joints of the injured hemostatically normal mice (Figure 3A). Nevertheless, a single dose of N9-GP 250 IU/kg given following injury provided protection that was superior to a single dose of unmodified rFIX 250 IU/kg (P = .0111) and which was at least as good as the protection afforded by rFIX dosed according to the 2 repeated-dose regimens.

Consistent with the time course described previously, the cartilage of NS placebo-treated FIX^−/− mice demonstrated more apoptotic chondrocytes and a lower mean chondrocyte number than cartilage of NS-treated hemostatically normal mice at 2 weeks after injury (Figure 3B-C). All FIX replacement regimens provided some chondrocyte protection. A trend toward increased chondrocyte cell death in the mice that received a single dose of rFIX was not, however, significantly different from WT mice and not associated with significantly diminished chondrocyte counts at 2 weeks following injury. A single dose of N9-GP or a 3- or 8-dose regimen of rFIX resulted in protection from chondrocyte apoptosis to a level similar to injured WT mice.

The heavy iron staining observed at 2 weeks after hemarthrosis in hemophilic mice (Figure 3D; supplemental Figure 4B) was improved by rFIX therapy (Figure 3D). A single dose of rFIX did not prevent the iron deposition being statistically greater than that observed in hemostatically normal mice. Conversely, intensified repetitive-dose rFIX regimen or a single dose of N9-GP prevented iron deposition in the joints to a level that was not statistically distinguishable from that observed in hemostatically normal mice (Figure 3D; supplemental Figure 4B). Similar patterns of wound healing were observed in regard to synovial macrophage residence and neoangiogenesis, with single-dose N9-GP resulting in healing similar to multidose rFIX therapy (Figure 3E-F).

We have previously reported that hemophilia A mice, examined by microCT scan, exhibit ectopic mineralization and roughening of the surface of the articulating bones following hemarthrosis.²⁶  Hemophilia B mice in this study exhibited these same acute changes involving distal femur, proximal tibia, and patella (Figure 4A). Following the induced hemarthrosis, joints from animals treated with N9-GP demonstrated preservation of articular surface smoothness that was not statistically significantly different from that seen in hemostatically normal WT mice. rFIX was less effective at preserving the integrity of the surface of the articulating bones (Figure 4B).

The untreated hemophilia B mice in this study also experienced an acute loss of greater than one-third of the volumetric bone mineral density (vBMD) measured at the proximal tibia adjacent to the blood-exposed knee when compared with hemostatically normal mice (Figure 5A-B). The vBMD measures the average mineral density over the entire trabecular compartment, including the space between the bone tissue, and so this microCT measurement is similar to that of a clinical dual-energy X-ray densitometry bone density scan but performed in 3-dimensional space. Mice given rFIX for 1 to 4 days after wounding exhibited diminished BMD, which was, however, normalized by 8-dose rFIX treatment during the 2 weeks of healing. Loss of BMD was not observed in mice treated with single-dose N9-GP following hemarthrosis.

The trabecular connectivity density, separation, and number are all structural microarchitectural parameters that may correlate with bone strength. At 2 weeks postinjury, in comparison with WT animals, untreated FIX^−/− group displayed a 28% increase in trabecular spacing (supplemental Table 2), consistent with diminished bone strength. Smaller trends regarding loss of trabecular number and connectivity density in FIX^−/− mice did not reach statistical significance. Normal trabecular bone properties were best preserved by rFIX D0-13 or N9-GP D0 replacement therapy.

The induced joint hemorrhage model in hemophilia mice has been widely used to examine histopathologic and molecular changes following hemophilic joint hemorrhage, as well as therapeutic interventions. We examined the hypothesis that joint wound healing following a single bleeding exposure is qualitatively abnormal in the absence of normal thrombin generation. We show that wound healing following bleeding into joints of hemophilia B mice is impaired. Most strikingly, the absence of normal hemostatic potential is associated with a diminished capacity to clear blood and heme iron from the joint space. The abnormal morphologic parameters seen here following hemophilic joint bleeding closely parallel those previously demonstrated in another hemophilic healing model of a dermal wound.^6,27,28 

Following the induced joint wound, the animals with normal hemostasis (with or without the additional applied blood load) demonstrate only modest inflammatory cellular, synovial proliferative, and angiogenic changes followed by orderly resolution. (Table 1; Figures 1B and 2; supplemental Figure 2). In the absence of coagulation FIX, the angiogenic and proliferative changes that occur are pathologic rather than physiologic in degree, and the resolution of these processes is delayed or incomplete at the end 8 weeks’ observation. The time course of neoangiogenesis and vascular remodeling within the FIX^−/− mouse joint that we describe are very consistent with changes described in a careful evaluation of the same hemarthrosis model performed in FVIII^−/− mice by Bhat et al.²⁹  Other processes that have no physiologic role in wound healing (eg, heme iron deposition, chondrocyte apoptosis) do not occur following wounding in the nonhemophilic animals (even when challenged with an applied blood load) but are prominent in hemophilia.

The current study demonstrates for the first time that induced joint bleeding in hemophilia B mice is associated with disrupted bone homeostasis, including diminished vBMD of trabecular bone and heterotopic mineralization of the external surface of the bones adjacent to hemarthrosis not seen in hemostatically normal mice that experience the same injury. These findings are consistent with our previous observations in hemophilia A mice.²⁶  The deficits observed following hemarthrosis may compound a congenital tendency to low BMD that has been described in hemophilic mice.^30,31  These animal findings are of particular interest in light of increasing awareness of the high prevalence of low BMD in individuals with hemophilia A and B.^32-38  In addition, high-resolution peripheral quantitative CT examination of an adult hemophilia population has recently correlated more severe hemophilic arthropathy with low BMD.³⁹ 

The consistency of the findings of abnormal wound healing between the hemophilic incisional wound model and the joint wound model suggests that studies of the pathophysiology and therapy of hemophilic joint disease that are performed in hemostatically normal animals should be interpreted with caution. Attempts to reproduce hemophilic arthropathy by modeling repeated joint bleedings in hemostatically normal dogs,²  or with repeated autologous intra-articular blood injections in nonhemophilic rats,³  led investigators to conclude that blood exposure alone could not lead to hemophilic joint damage.^2,3  This observation is strikingly different from the gross pathology that develops following spontaneous joint bleeding in hemophilic dogs,^40,41  hemophilic rats,⁴²  and following 3 induced joint hemorrhages in hemophilic mice.⁹  This is further corroborated by our observation that blood injected into the joint of WT animals is cleared efficiently.

Thrombin serves many functions beyond hemostasis, so that absent thrombin generation may contribute to abnormal joint wound healing via additional mechanisms beyond joint exposure to hemorrhage. During normal cutaneous wound healing, thrombin promotes the physiologic neoangiogenesis and fibroblast migration that are required for orderly remodeling. Thrombin and thrombin signaling through protease activated receptor 1 have been reported to enhance incisional wound healing.⁴³  The absence of thrombin may contribute to the disorganized and pathologic neoangiogenesis and fibroblast infiltration in the blood-exposed hemophilic joint. In the absence of thrombin, protein C activation is impaired; anti-inflammatory actions of activated protein C to protect the wounded joint would be lost and neutrophil infiltration unchecked.^44,45 

Absent thrombin generation in hemophilia, as well as the coincident loss of activated protein C activation by thrombin, translates to loss of osteoblast stimulation and unopposed inhibition of osteoblast apoptosis and may contribute to impaired bone remodeling after wounding.^46-49  Additionally, plasmin and fibrinolysis may be directly toxic to cartilage. Normal activation of thrombin-activatable fibrinolysis inhibitor results in diminished formation of plasmin, attenuation of fibrinolysis, increased clot stability, and anti-inflammatory effects that may protect the healing joint wound.^13,50,51  However, relatively high concentrations of thrombin are required to activate thrombin-activatable fibrinolysis inhibitor in the low tissue factor environment within the joint.⁵¹  In summary, absent thrombin generation in hemophilia potentially disrupts normal joint remodeling in synovium, cartilage, and bone by multiple mechanisms that compound the direct mechanical and toxic effects of hemorrhage.

A limitation of these studies is that the FIX^−/− mice rarely demonstrate spontaneous bleeding. There is, however, no evidence that the body has different mechanisms for healing traumatic hemarthroses when compared with spontaneous hemarthroses. Another possible limitation is the attempt to model the healing response to an autologous blood load in hemostatically normal mice. Exposure of the joint to citrated blood may not recapitulate the blood challenge that occurs during traumatic clinical intraarticular hemorrhage in individuals who do not have hemophilia. The volume of blood injected (5 μL) was chosen based on our work suggesting this amount of blood will fill the mouse joint space (data not shown) and demonstrates a clear difference in the clearance of blood from the joint in the presence (compared with the absence) of normal hemostasis. It is possible that the WT mouse joint challenged with a broad range of injected blood loads might demonstrate different healing outcomes, although it would be difficult to interpret the relative contributions of barotrauma, hypoxia, and differing load of blood components to the observed pathology.

Restoring hemostatic potential only until hemostasis was obtained did not result in optimal joint wound healing. Instead, optimal joint health is achieved in this model by extending clotting factor activity during wound healing.^27,28,52 

The use of N9-GP, which is an FIX engineered for prolonged survival in circulation, resulted in improved wound healing when compared with single or repeated doses of rFIX. Extending hemostatic activity during wound healing by single-dose administration of N9-GP resulted in healing that was statistically not different from the healing seen in WT mice in regard to iron deposition and BMD. Mice that received rFIX administered in the “enhanced episodic” schedule of 3 doses of rFIX nevertheless healed less well than WT mice with regard to bone health, experiencing diminished vBMD and increased surface roughening. rFIX infused throughout 2 weeks of healing resulted in correction of vBMD (articular surface irregularity was not prevented). Notably, mice treated with N9-GP on day 0 and day 7 demonstrated no significant improvement in any parameters when compared with mice that received N9-GP on day 0 alone (supplemental Figure 4). The pharmacokinetics and pharmacodynamics of a single dose of N9-GP in hemophilia B mice given at the dose used in this study are well established. This dose provides circulating activity (>5% to 10% of normal) and clotting during bleeding and thrombotic challenges for at least 5 to 7 days,²²  which appears to be the critical period for optimal joint and bone outcomes in this gross hemarthrosis model. We are currently investigating the possibility that prophylactic rFIX or FIX engineered to have extended circulating kinetics (N9-GP) can achieve fully normal joint healing following hemarthrosis that is equivalent to healing in hemostatically normal mice.

The authors would like to acknowledge the outstanding technical assistance, consultation, and discussion of Elizabeth P. Chappell, Jacqueline B. Mickelson, Anthony G. Lau, and Tom Knudsen.

This work was supported in part by research funding from Novo Nordisk (D.M.M. and P.E.M.).

Contribution: J.S. performed the studies, generated and adapted “Methods,” prepared the figures, and reviewed and revised the manuscript; B.H. performed the studies and reviewed and revised the manuscript; E.W.L. performed the studies, adapted “Methods,” prepared the figures, and reviewed and revised the manuscript; S.T. reviewed and revised the manuscript, prepared the figures, and performed statistical analysis; P.B.J. performed statistical analysis; M.H. and M.E. provided a critical review of the study design and edited the manuscript; D.M.M. provided a critical review of the study design, assisted the studies, and edited the manuscript; T.A.B. designed and oversaw the studies, reviewed and interpreted the data, and edited the manuscript; and P.E.M. designed and oversaw all studies, interpreted all data, and wrote the manuscript.

Conflict-of-interest disclosure: P.E.M. during the conduct of these studies received research support through the University of North Carolina from Asklepios BioPharmaceutical and Novo Nordisk. He has received research support in the past from Baxter Healthcare, Novo Nordisk, Pfizer, and Prolor. He holds patents licensed to Asklepios, for which he receives royalties. He has received payment for consultation, services, and for speaking for Asklepios, Chatham LLC, Baxter Healthcare, and Pfizer and has additionally consulted for Bayer, Novo Nordisk, and Biogen. Following the completion of these studies, but during a portion of the time of manuscript preparation, he has been an employee of Baxalta (now Shire). B.H. has received research support from the Novo Nordisk Haemophilia Research Fund following his active participation in the reported research. M.H. reports receiving grants from Novo Nordisk during the conduct of the study and grants from CSL Behring and Boehringer Ingelheim outside the submitted work. M.E. is an employee of Novo Nordisk. D.M.M. reports receiving grants, personal fees, and nonfinancial support from Novo Nordisk during the conduct of the study and grants from Novo Nordisk outside the submitted work. T.A.B. has received remuneration for consulting and speaker work for Baxter Health Care Corporation and research support from Novo Nordisk. The remaining authors declare no competing financial interests.

Correspondence: Paul E. Monahan, University of North Carolina, Chapel Hill, 141 Graylyn Dr, Chapel Hill, NC 27516; e-mail: pablomonoloco@gmail.com.