Flowchart of progress of the corneas and recipients. Group 1 corneas were stored from 5 to 12 days; group 2 corneas, 21 to 35 days. Three corneas allocated to group 2 were discarded owing to excessive endothelial cell loss (n = 2) or bacterial contamination (n = 1). The number of paired recipients was thus reduced to 25 in each group for the statistical analysis.
The biexponential model fitted to the loss of endothelial cells during the first 12 months in groups 1 and 2. Group 1 corneas were stored from 5 to 12 days; group 2 corneas, 21 to 35 days. The equations were ECD = 1973e−0.0068t+ 578e−3.2t for group 1 and ECD = 1747e−0.0078t+ 531e−3.6t for group 2, in which ECD indicates endothelial cell density; e, exponential function; and t, postoperative time in months. The half-times of the slow component of the decline were 102 and 89 months for groups 1 and 2, respectively.
Long-term postoperative decline in endothelial cell density (ECD) for initial donor ECDs of 2551 cells/mm2 (group 1) and 2278 cells/mm2 (group 2). Group 1 corneas were stored from 5 to 12 days; group 2 corneas, 21 to 35 days. The percentage fall in ECD with time was predicted using the equation ECD = 1397e−0.0027t + 1458e−0.08t and adapted to the initial densities of the 2 groups. (In the equation, e indicates exponential function and t, postoperative time in months.) This led to ECD = 1248e−0.0027t + 1303e−0.08t for group 1 and ECD = 1115e−0.0027t + 1163e−0.08t for group 2. Threshold ECDs of 500 and 250 cells/mm2 are shown. The time needed for the ECD to decline to these values would be 28 and 50 years, respectively, for group 1, and 25 and 46 years, respectively, for group 2.
Thuret G, Chiquet C, Bernal F, Acquart S, Romanet J, Mouillon M, Hegelhoffer H, Burillon C, Damour O, Maugery J, Armitage WJ, Gain P. Prospective, Randomized Clinical and Endothelial Evaluation of 2 Storage Times for Corneal Donor Tissue in Organ Culture at 31°C. Arch Ophthalmol. 2003;121(4):442-450. doi:10.1001/archopht.121.4.442
To compare the endothelial and clinical outcome of penetrating keratoplasty with corneas stored in organ culture for up to 12 days (5-12 days; group 1) or more than 21 days (21-24 days; group 2).
We conducted a controlled double-masked trial. Storage durations were randomly assigned to the paired corneas, and endothelial cell density (ECD) was measured at the start and end of organ culture. Patients with a low rejection risk and preoperative ECD within the reference range were randomly assigned to 1 of the 2 groups and underwent an 8.25-mm penetrating keratoplasty (n= 25 pairs). Follow-up at day 5 and months 1, 6, and 12 included central ECD, morphometry, graft transparency, visual acuity, pachymetry, and complications. The main outcome measure was the central ECD at month 12.
At the end of organ culture, ECD of the group 1 corneas was higher by 273 cells/mm2 (95% confidence interval [CI], 178-368; P<.001). One year after penetrating keratoplasty, the group 1 ECD was still comparably higher by 227 cells/mm2 (95% CI, 43-411; P = .02). Graft transparency, pachymetry, and complication rate did not differ at any time. In group 1, visual acuity was better at month 1.
Shorter organ culture allows delivery of corneas with higher ECD. Recipients with ECD within the reference range and low rejection risk retain this initial benefit 1 year postoperatively. The higher endothelial cell capital may prevent or delay late endothelial failure, the leading cause of graft failure in these recipients. We therefore prefer short-term storage for such recipients.
CORNEA SURVIVAL after penetrating keratoplasty (PKP) depends on many factors. The most important of these is preoperative corneal disease in the recipient, which influences rejection risk and postoperative endothelial cell loss.1,2 The number of endothelial cells provided by the graft also affects survival to some extent, especially in recipients with normal endothelial cell density (ECD) and low rejection risk. Several recent studies have suggested that late endothelial failure can be avoided if the graft provides the highest possible number of viable endothelial cells.3,4
Cornea preservation in organ culture causes endothelial cell loss, which increases with storage time.5 Therefore, the simplest way to graft corneas with a high ECD is to shorten cornea storage. However, no study has formally demonstrated the influence of organ culture time on long-term cornea survival or even medium-term central ECD in recipients. The literature on organ culture and the factors responsible for post-PKP cell loss consists mainly of retrospective cohort studies. The absence of cornea pairing and the heterogeneous nature of the recipients (eg, high or low ECD, high or low rejection risk) and surgical procedures (eg, single or combined PKPs of varying diameters)6- 12 generate a high background noise, which obscures the true influence of organ culture time on the outcome of interest. One retrospective cohort study conducted on homogeneous recipients with a high ECD and low rejection risk (keratoconus only)12 did not demonstrate the influence of storage time on postoperative endothelial cell loss because of a lack of statistical power. The only reported prospective, controlled trial that compared 2 organ culture times (14 vs 30 days) did not highlight any harmful effect of prolonged storage, 13 but in that study the corneas were not paired, the recipients were heterogeneous, and the primary end point was graft transparency at 18 months rather than ECD. A checklist for randomized, controlled trials is provided in Table 1.
In the absence of consensus, most European ophthalmologists perform grafts in their patients without preference for a given organ culture time. To resolve this crucial issue, we conducted a double-masked, randomized, prospective controlled trial of endothelial survival 1 year after PKP in homogeneous recipients(ie, with a high ECD and low rejection risk). We compared 2 groups of paired corneas differing only by organ culture time. The purpose of this study was to determine whether to recommend a reduction in organ culture time for corneas in this type of recipient. As pseudophakic bullous edema becomes rarer, such recipients, especially those with keratoconus, are now the main candidates for grafting.14,15
We conducted a double-masked, 2-center, prospective comparative trial, in which storage time of paired corneas and their allocation to recipients were randomized. The institutional review boards of both centers approved the project. Participants signed a statement of informed consent approved by the regional ethics committee. When a pair of corneas arrived in 1 of the 2 banks, organ culture time was randomly assigned. One cornea of the pair was stored from 5 to 12 days (short-term storage); the other, 21 to 35 days(long-term storage). Corneas were collected after in situ corneoscleral excision in accordance with statutory French practice16 and immediately placed in Inosol organ culture medium (Bausch & Lomb–Chauvin–Opsia, Labège, France) at 31°C. All cornea media were renewed at the start of storage at day 2 to day 4, then at day 14 for the group 2 corneas. Two days before delivery, corneas were placed in Exosol deswelling medium (Bausch& Lomb–Chauvin–Opsia) containing 5% dextran T500. Optimized bacteriological and fungal safety tests were performed at the start (day 2 to 4) and end (2 days before delivery) of organ culture using blood culture bottles and automated detection of contamination, as previously described.17 Central ECD was measured at the same moment under a light microscope after dilation of intercellular spaces using 0.9% sodium chloride and a fixed-frame counting technique; percentage of cell death was measured using 0.4% trypan blue. Given that cell loss is inevitable during prolonged organ culture, the cornea pairs were included only if the initial ECD of each cornea was greater than 2400 cells/mm2. For each cornea we counted the percentage of total endothelial cell loss during storage, corresponding to 100 × (initial ECD − final ECD)/initial ECD, and the percentage of daily endothelial cell loss corresponding to the percentage of total endothelial cell loss divided by the number of days of organ culture. Corneas were discarded when microbiological test results were positive or when, at the end of storage, they displayed an ECD of less than 2000 cells/mm2.
For the recipients, eligibility criteria included stage 4 keratoconus but without prior hydrops, lattice dystrophies with impairment of the posterior stroma, and a normal ECD preoperatively, verified by specular microscopy. Exclusion criteria included previous intraocular surgery. Eligible subjects were randomized by center into group 1, which received the cornea undergoing short-term storage, and group 2, which received the cornea undergoing long-term storage. A randomization list, drawn up from a table of random numbers, was provided in each center to an independent person (the chief nurse of the ophthalmology department), who ensured that randomization was centralized and who did not take part in cornea examination or patient follow-up. If a cornea was discarded at the time of the endothelial or microbiological tests performed during organ culture, the planned recipient took the following place in the waiting list without leaving his or her randomization group. In the event of patient dropout, the recipient of the paired cornea was excluded from the comparisons.
The 50 single PKPs were performed by 4 experienced surgeons (J.P.R., M.M., J.M., and P.G.) (each had performed more than 500 grafts) using the following identical protocol: general anesthesia, 8.25-mm-diameter trephination(punching) of the donor from the endothelial side with a vacuum donor cornea punch (Barron; Katena Products, Inc, Denville, NJ), 8.00-mm-diameter trephination of the recipient by means of a radial vacuum trephine (Barron; Katena Products, Inc), running suture with 10/0 nylon monofilament, protection of the endothelium by means of viscoelastic substance, and subconjunctival injection of neomycin sulfate (0.35 MIU/100 mL) and dexamethasone phosphate. We recorded the operating time and noted any incidents. The surgeons performed the grafting procedures without knowing the storage time. During PKP, the surgeon graded the quality of the tissue using a 3-level score consisting of excellent for perfect visibility of details of the iris, clear stroma, and few or no endothelial folds; good for imperfect visibility of details of the iris, minimal stromal edema, and moderate endothelial folds; and acceptable for poor visibility of details of the iris, considerable stromal edema, and numerous endothelial folds.
Postoperative treatment consisted of application of topical norfloxacin and tropicamide morning and evening until reepithelialization of the graft, and then topical corticosteroids 3 times daily in tapered dosages during the first 9 postoperative months. In the event of elevated intraocular pressure(IOP) of greater than 20 mmHg or endothelial rejection, the same therapeutic protocol was followed in each center. Briefly, elevated IOP was treated with topical timolol maleate if it remained greater than 20 mmHg after reduction of corticosteroid dosage. Rejection episodes were treated with oral prednisone in tapered dosages during 2 months, topical dexamethasone phosphate (6 times per day in tapered dosages during 2 months), and daily subconjunctival injection of dexamethasone phosphate for 5 days.
The patients were then followed up by 2 ophthalmologists (1 per center[G.T. and F.B.]), who were masked to the cornea storage time. Patients were examined at 5 days and 1, 6, and 12 months postoperatively. On each visit, the following 6 variables were measured: graft transparency, best corrected visual acuity (BCVA), astigmatism, pachymetry, central ECD, and endothelial morphometry. Graft transparency was scored according to the following 3 levels: very clear if the stroma was clear, with few or no endothelial folds; clear if the stroma was clear but with numerous endothelial folds, or if the stroma was slightly edematous and had few endothelial folds; and nebulous if the stroma was slightly edematous with numerous endothelial folds, or if the stroma was highly edematous whatever the aspect of the endothelium. We determined BCVA with spectacles and without contact lenses. Astigmatism was measured using the Javal ophthalmometer. We obtained corneal pachymetry with the mean of 10 measurements at the graft apex (AL 1000; Tomey GmbH, Erlangen, Germany). Central ECD and endothelial morphometry were determined using noncontact specular microscopy (SP2000; Topcon Corporation, Tokyo, Japan) aided by the Imagenet 2000 version 2.03 software program (Topcon Corporation). Measurements were taken on the widest surface area allowed by the software on at least 75 cells to ensure reliable results, as shown using noncontact specular technology.18 In addition, to prevent any error arising from the automatic analysis method, 19 the cell contours were systematically traced by hand. All analyses were then performed by a single investigator, always masked to storage time.
The primary end point was central cornea ECD in recipients at 1 year, measured by specular microscopy. Endothelial cell density was deemed a good surrogate criterion for graft survival. It is an intermediate criterion, easy to measure reliably and reproducibly, and its relation to graft survival has been widely demonstrated.3,20,21 The sample size of 21 corneas in each group would have been large enough to detect a statistical significance (α = .05; β = .1), assuming an expected size difference at 1 year after the graft of 300 cells/mm2 and an SD of 300 cells/mm2. This SD was estimated according to the data of teams using organ culture.9,22 Given the risk for an intercurrent event (postoperative trauma) and patient dropout, we enrolled 28 patients in each group.
Unless otherwise indicated, data are expressed as mean ± SD. For paired data, we compared the normally distributed quantitative variables using a 2-tailed paired t test and the qualitative variables using the MacNemar nonparametric paired test. For the ECDs, to keep statistical tests to a minimum, and because of some patients' nonattendance of intermediate visits, we made comparisons only at the beginning and end of organ culture, and at month 12 in recipients (primary end point). For nonpaired data, rates of rejection and raised IOP in the 2 groups were compared using a Fisher exact test, and endothelial cell loss in the 2 groups at month 12 was compared with the full recipient set using a Mann-Whitney test. We performed statistical analysis using SPSS software, version 10.0 for Windows (SPSS Inc, Chicago, Ill), and P<.05 was considered significant.
The characteristics of the 2 recipient groups, presented in Table 2, were not statistically different. Three corneas allocated to group 2 recipients were discarded during storage, reducing the number of recipient pairs to 25. This did not influence comparability of the 2 groups (data not shown). A flowchart (Figure 1) shows the progress of corneas and recipients during the study.
We included 56 corneas (28 pairs) collected from 0 (heart-beating donors) to 27 hours post mortem (mean, 10 hours; SD, 8 hours). Non–heart-beating donors accounted for 19 (68%) of 28 pairs of corneas (18 men and 10 women). Mean donor age was 48.1 ± 20.4 years.
Three corneas in the long-storage group 2 (11%) were lost during organ culture, one through bacterial contamination and the others through excessive endothelial cell loss. No corneas were lost in group 1. The characteristics of the 3 discarded corneas and of their fellow grafts are presented in Table 3.
The characteristics of the 25 remaining pairs of corneas are presented in Table 4. The ECDs in both groups at the start of organ culture were comparable. The group 2 corneas, whose mean storage time was 25.3 ± 4.7 days (range, 21-34 days), lost more endothelial cells (16.2% ± 5.2%) than those in group 1 (6.4% ± 4.8% [P<.001]), whose mean storage was 9.1 ± 1.9 days (range, 5-12 days). At cornea delivery, this led in group 1 to an ECD higher by 273 of 2551 cells/mm2 (95% confidence interval [CI], 178-368 [P<.001]), corresponding to 10.7% extra cells.
During surgery, the tissue quality score for both groups was comparable. It was judged excellent for 21 (84%) of the 25 group 1 corneas and 20 (80%) of 25 group 2 corneas (P>.99). All the operations were incident free, and their mean durations in groups 1 and 2 were comparable at 41 ± 8 vs 41 ± 6 minutes, respectively (P = .93).
At 12 months postoperatively, we found a significant difference in ECDs of 227 of 1818 cells/mm2 (95% CI, 43-411 [P =.02]), corresponding to 12.5% extra cells for the recipients in group 1. This figure was comparable to the difference at delivery. The percentage of endothelial cell loss at 1 year, calculated by the formula 100 × (ECD at delivery− ECD at month 12)/ECD at delivery, was not statistically different between groups 1 and 2, at 29.9% (95% CI, 24.6%-35.3%) and 31.3% (95% CI, 25.4%-37.2%), respectively (P = .73). Overall, cell loss was 30.6% (95% CI, 26.8%-34.5%).
At the intermediate stages, there seemed always to be a difference in corneal ECD in favor of the group 1 recipients, in whom ECD was stable throughout follow-up (234 and 237 cells at months 1 and 6, respectively).
Although ECD was measured on a smaller number of cells than in other counts (mean, 49 ± 27 vs 102 ± 25 cells for all others [P<.001]), high cell loss of 17.5% (95% CI, 13.1%-21.9%) was suspected in both groups combined at day 5.
No recipient experienced an infectious complication. Six corneas (12%) soon presented a fully reversible endothelial rejection (n = 4 in group 1 and n = 2 in group 2 [P = .67]). At month 12, these 6 patients had a mean endothelial cell loss of 42.3% ± 19.6%, which was not significantly different from that of the 44 recipients without rejection(P = .15). Four patients (8%) experienced elevated IOP that resolved spontaneously or normalized under treatment (n = 3 in group 1 and n = 1 in group 2 [P = .61]). At month 12, these 4 patients had a mean endothelial cell loss of 35.6% ± 25.9%, which was not statistically different from that of the 46 recipients without elevated IOP (P = .61).
All of the corneas were transparent at the end of the study. With the exception of BCVA at month 1, which seemed slightly better in group 1, no difference in BCVA between the 2 groups could be highlighted at month 6 or 12 and no differences in corneal transparency, astigmatism, pachymetry, or endothelial cell morphometry throughout follow-up (Table 5).
This randomized, prospective study of paired corneas shows that 1 year after PKP, in recipients with preoperative ECD within the reference range and a low rejection risk, a difference in ECD of 227 cells/mm2 remains corresponding to an extra endothelial cell capital of nearly 12% for the recipients of corneas undergoing short-term storage in organ culture. Because the study looked at paired corneas, and because the 2 cell decrease curves were parallel, this difference could be attributed only to the mean 16.2-day organ culture extension for group 2. At delivery, the group 1 corneas (mean storage time, 9.1 days) presented an ECD higher by 273 cells/mm2 than group 2 corneas (mean storage time, 25.3 days) and retained this benefit throughout follow-up.
However, throughout follow-up, we highlighted no difference in the other functional (transparency, BCVA, astigmatism, pachymetry) or morphological variables (coefficient of the variation of the cell surface and endothelial hexagonality). Only BCVA at month 1 seemed better in group 1. However, this markedly faster visual recovery must be considered with caution in view of the large number of statistical tests used to analyze this secondary criterion. Moreover, no variable (pachymetry or the transparency score) explained this result.
The benefit of the highest possible ECD in the first year after PKP is obvious. The extra endothelial cell reserve, especially in this population, allows patients to cope better with events liable to affect the endothelium, in particular elevated IOP and endothelial rejection. Both occur most often in the first year after grafting.23- 28 As Bourne et al7 emphasize, the number of endothelial cells surviving on corneal transplants is important for long-term graft survival because grafts with more cells probably have a greater capacity to withstand the effects of aging, inflammation, and rejection episodes. Although our as yet limited perspective prevents a formal conclusion, this extra cell reserve is likely to prolong graft survival into the very long term by reducing late endothelial failure. In this young population with a low rejection risk, late endothelial failure is one of the main causes of late graft failure.3,20,21 Several studies of all-comer recipients, comprising high and low preoperative ECD and variable rejection risks, have shown that an early deficit of 500 cells/mm2 is associated with a significant rise in late endothelial failure risk.3,20,21 Of a selected population of recipients such as ours, a smaller difference of about 250 cells/mm2 probably would have similar effects. To predict late endothelial failure, Redmond et al9 fitted a monoexponential decay curve to the postoperative loss of endothelial cells from grafts for 4 years in all-comer recipients. This model, however, greatly overestimates cell loss in the longer term. Adapting a more complex model based on the sum of 2 exponentials(e), ECDt = ae−pt+ be−qt(W.J.A., unpublished data; ECDt is cell density at time t [postoperative time in months], p is equal to 0.693/half-life of the first curve component, q is equal to 0.693/half-life of the second curve componenet, a and b are constants whose sum is equal to the initial ECD), to our data allows the percentage of the loss of cells with increasing postoperative time to be estimated. The model was fitted separately to the short- and long-term storage groups (Figure 2). The rate of decline in ECD during the 12-month follow-up was very similar in both groups. Given that these equations were derived from such a limited follow-up period, it would not be justified to extrapolate them into the long term. An equation that had been derived from a data set of long-term measurements of ECD was therefore used to determine the percentage of fall in ECD with time (Figure 3).This allowed the change in ECD with time to be calculated for initial donor ECDs of 2551 (group 1) and 2278 cells/mm2 (group 2). These curves allowed estimation of the long-term survival times, assuming different endothelial decompensation thresholds. The ECD threshold below which the endothelium can no longer ensure corneal deturgescence is difficult to define because it varies according to the circumstances of endothelial loss.29 Different authors have reported decompensation thresholds of 250, 30 300, 31 400, 32,33 or 50034 cells/mm2. Figure 3 shows the predicted decline in ECD in the 2 groups and the time taken to reach ECD thresholds of 500 and 250 cells/mm2.If decompensation is expected to occur at an ECD of 500 cells/mm2, graft survival in groups 1 and 2 would be 28 and 25 years, respectively. A lower decompensation threshold of 250 cells/mm2 is probably more realistic, based on personal observation by means of specular microscopy of numerous keratoconi grafted more than 30 years ago with a still-transparent cornea (J.M., unpublished data, 2000). If decompensation did not occur until an ECD of 250 cells/mm2 had been reached, graft survival would be 50 and 46 years in groups 1 and 2, respectively. Whether these differences would be considered clinically significant in grafts that may be predicted to last for more than 25 years is a difficult question, but one that needs to be addressed.
However, our results will not necessarily be relevant to all-comer recipients. The initial benefit of 273 cells/mm2 at delivery would probably be quickly lost by recipients with a low preoperative ECD, such as those with Fuchs dystrophy or bullous keratopathies, or by recipients at high rejection risk who in the first postoperative year presented with cell losses of far greater than our patients.35- 37
Our groups showed a substantial early endothelial cell loss of 17.5% at day 5. This massive cell loss has been noted by other authors, ie, 17% at 4 days38 or 23% at 1 week39 after PKP, but in all-comer recipients. This similarity suggests that this very early cell loss depends little on the recipient conditions, unlike later cell loss. It depends on multiple factors. First, it is increased by a methodological bias. The actual ECD at delivery is overestimated because it is measured several days before grafting, while in hypothermic storage or in organ culture. In organ culture, placement of the cornea into deswelling medium 48 hours before grafting is responsible for significant additional endothelial cell loss, estimated at 8.7% by Borderie et al.40 We showed in a previous experimental work that the viability test performed during organ culture, using trypan blue, largely underestimated the number of endothelial cells engaged in an irreversible process of cell death.41 Second, the trauma of mechanical trephination creates a cellular defect 150 to 200µm wide, 42 which for an 8.25-mm-diameter trephination, as in our series, represents endothelial cell loss of 7% to 9.5%. Third, the graft suture, although performed only by experienced operators in our study and under a viscoelastic protection, is probably harmful. Fourth, the role of the immunological reaction immediately after the operation, even in this population with low rejection risk, is certainly not negligible.
Some authors suggest that rejection frequency rises in cases in which organ culture is less than 8 days.10 This is partly explained by the initial presence of antigen-presenting cells, which disappear after the first week of organ culture.43- 45 We found no difference between our groups in incidence of rejection. Although we remain cautious, as this was not the main focus of our study, this result is not surprising. Group 1 patients who experienced a rejection had received corneas stored from 8 to 12 days, which is above the threshold proposed by these authors. Moreover, a higher rejection rate has never been demonstrated with corneas stored at 4°C or with fresh corneas. Lastly, and more generally, no previous study found any influence of storage variables on rejection incidence.6,20,21,25,28,46
Organ culture was originally designed with the following dual purpose: to allow prolonged preservation of endothelial viability, and to ensure maximum microbiological safety of corneas.13,47- 49 Although in practice prolonged organ-culture time rarely exceeds 20 to 25 days, as testified by the 2002 register of the European Eye Bank Association, 50 to some extent this practice allows selection of the corneas with the most resistant endothelium by means of a stress test.5,51 However, this causes elimination of corneas that would have been acceptable for clinical use, as we observed in our series, in which 3 group 1 corneas were grafted without incident and with comparable evolution, whereas their fellows were discarded. This loss of corneas is due to contamination or to a technical error during media renewal (1 case in our series), or, more often, to endothelial loss of about 0.7% per day of additional organ culture, which caused the ECD to fall below the delivery threshold of 2000 cells/mm2.
The main advantage of organ culture is to allow delivery of corneas with very low contaminating potential.52 This requires a typical quarantine period of 10 to 12 days, 5,50,53 and depends in practice on the sensitivity and detection speed of the microbiological tests used by the banks. In previous studies, we showed that this quarantaine can be shortened by using blood bottles at the start and end of organ culture, as these are far more sensitive in detecting bacteria and fungi than conventional methods.17,54 In the present series, no infection occurred in the group of patients who received corneas undergoing short-term storage (13 corneas were stored for 5-8 days) that were microbiologically tested using this effective method.
Our study shows an endothelial benefit persisting 1 year after PKP in recipients with high ECD and low rejection risk, who have received a cornea undergoing short-term storage in organ culture. The very long-term follow-up of the 2 cohorts of patients will show whether prolonging organ culture by several days causes the loss of several years of graft survival. Pending that finding, we consider it legitimate to prefer the shortest possible storage of corneas allocated to such recipients.
Corresponding author: Gilles Thuret, MD, Department of Ophthalmology, University Hospital, 42055 Saint-Etienne, France (e-mail: firstname.lastname@example.org).
Submitted for publication August 13, 2002; final revision received October 18, 2002; accepted December 20, 2002.
This study was supported by the Department of Clinical Research, Saint-Etienne University Hospital, Saint-Etienne, France (Projet Hospitalier de Recherche Clinique national 9801068).
We thank Hervé Décousus, MD, and Sylvie Laporte, PhD, for help in designing this randomized controlled trial.