Published ahead of print on April 19, 2007, doi:10.1164/rccm.200608-1238OC
© 2007 American Thoracic Society doi: 10.1164/rccm.200608-1238OC
Phase 2 Randomized Safety and Efficacy Trial of Nebulized Denufosol Tetrasodium in Cystic Fibrosis1 The Children's Hospital, Denver, Colorado; 2 University of Colorado, Denver, Colorado; 3 University of North Carolina, Department of Biostatistics, Chapel Hill, North Carolina; 4 Inspire Pharmaceuticals, Inc., Durham, North Carolina; 5 Cystic Fibrosis Therapeutics Development Network, Seattle, Washington; 6 Cincinnati Children's Hospital, Cincinnati, Ohio; and 7 University of Cincinnati, College of Medicine, Cincinnati, Ohio Correspondence and requests for reprints should be addressed to Robin Deterding, M.D., The Children's Hospital, 1056 East 19th Avenue, B395, Denver, CO 80218. E-mail: deterding.robin{at}tchden.org
Rationale: Denufosol tetrasodium is a selective P2Y2 agonist that enhances mucosal hydration and mucus clearance by activating Cl– secretion and inhibiting epithelial Na+ transport through a non–cystic fibrosis transmembrane conductance regulator mechanism in the lung. Objectives: To examine the safety and efficacy of 28 days of treatment with denufosol compared with placebo in patients with mild cystic fibrosis.
Methods: The study was a randomized, double-blind, multi-center, 28-day, phase 2 clinical trial of denufosol tetrasodium inhalation solution (20, 40, or 60 mg) versus placebo (normal saline). Patients with screening FEV1 Measurements and Main Results: Eighty-nine patients were randomized and received the study drug, 94% completed the study, and 98% were compliant with dosing. All treatments were generally well tolerated, with no dose–response trends observed with respect to safety parameters. The most common adverse event was cough (52% of placebo patients and 47% of denufosol patients). Five patients discontinued early due to adverse events, two on placebo and three on denufosol. Denufosol patients (pooling active doses) had significantly higher changes from baseline in FEV1 (P = 0.006), FEF25%-75% (P = 0.008), FVC (P = 0.022), and FEV1/FVC (P = 0.047) than placebo patients at the end of the study. Conclusions: Denufosol administered three times daily for 28 days appeared to be safe and well tolerated in this population with mild cystic fibrosis and provided preliminary evidence of potential benefit in lung function.
Key Words: clinical trial • lung function • adverse event
Cystic fibrosis (CF) is a recessive genetic disease (1) that affects approximately 30,000 Americans (2). Most deaths of patients with CF occur as a consequence of pulmonary disease. The disease is caused by mutations in the cystic fibrosis transmembrane regulator (CFTR) gene, which encodes for an apical membrane epithelial protein that functions as a cyclic AMP–regulated chloride channel and a regulator for other channels (3). Defective CFTR results in abnormal ion transport and reduced volume of airway surface liquid, which inhibits mucociliary and cough clearance and leads to chronic infection of the respiratory tract (4). Although survival has recently increased with the advent of better treatments (median predicted survival = 37 yr [2]), patients have significant morbidity (5). Agents that correct the underlying ion transport defects in the airways may prove useful in normalizing airway secretions, leading to improved mucociliary clearance, and prevention of chronic lung infections and progressive lung damage. Endogenous P2Y2 agonists regulate key airway host defense mechanisms. The P2Y2 receptor is abundant on the luminal surface of polarized epithelia lining bronchial surfaces (6). There are multiple effects of P2Y2 agonists on airway epithelial function, including stimulation of serosal to mucosal chloride and fluid transport by both inhibiting sodium absorption and stimulating chloride secretion (7–11); enhancement of mucin secretion from goblet cells, which promotes the trapping of foreign particles (12, 13); stimulation of cilia beat frequency (14); and promotion of surfactant release from type II alveolar cells (15). Denufosol tetrasodium [P1-(uridine 5')-P4-(2'-deoxycytidine 5')tetraphosphate, tetrasodium salt] is a novel P2Y2 agonist with a pharmacologic profile that parallels UTP and diquafosol (6, 16–18). UTP activates Cl– secretion and enhances mucociliary clearance in normal subjects and smokers (19–21). Denufosol is metabolically more stable than previously tested P2Y2 agonists (6- and 50-fold, respectively) and may have long-lasting effects in the airways of patients with CF, as increased tracheal mucosal velocity has been observed for up to 8 hours in sheep (6). An earlier phase 1 study showed that doses of 200 mg and above in nonsmokers (n = 40) and 100 mg and above in smokers (n = 40) were associated with mild cough, which was typically productive (22). A phase 1/phase 2 dose-escalation study showed that dose levels of 10, 20, 40, and 60 mg were generally well tolerated by patients with CF (23). The objectives of this study were to evaluate the safety and efficacy of three dose levels of denufosol compared with placebo administered three times daily over 28 days in patients with CF. Other objectives were to explore the utility of parameters measured from high-resolution computed tomography (HRCT) scans of the lungs as endpoints in CF clinical trials and to explore evidence of pharmacologic activity of denufosol as indexed by the amount of sputum expectoration over time. Some results of this study were previously reported in abstract form (24–26).
The study was a multi-center, double-blind, randomized, placebo-controlled, parallel-group trial of denufosol versus placebo administered three times daily via the Pari LC Star nebulizer (PARI Respiratory Equipment, Inc., Midlothian, VA). Patients were randomized in a 1:1:1:1 ratio to receive placebo, 20, 40, or 60 mg denufosol (0.9% wt/vol of sterile sodium chloride, 5, 10, and 15 mg/ml, respectively) for 28 days (Figure 1). The study was conducted through the Cystic Fibrosis Foundation (CFF) Therapeutics Development Network (TDN) at 14 CF centers, including 6 centers external to the TDN. Two planned interim analyses were performed to monitor the safety and tolerability. No changes were made to the protocol as a result of the Data Monitoring Committee reviews. Central randomization procedures were implemented via a web-based system across clinical centers. Study patients and all study personnel at the clinics and at the sponsoring organizations were blinded to treatment group assignments for the duration of the study.
Study Population Patients (aged 8–50 yr) were eligible if they had a confirmed diagnosis of CF by sweat chloride and/or genotype, an FEV1 greater than or equal to 75% of predicted normal based on Knudson standards (27), and an oxyhemoglobin saturation of greater than or equal to 90% on room air. Patients were required to be clinically stable and able to reproducibly perform spirometry. In addition, the Day 1 predose FEV1 had to be within ± 12% of the most recent screening FEV1.
Key exclusion criteria included abnormal renal or liver function (serum creatinine
Safety and Efficacy Parameters
Statistical Analysis The primary efficacy analyses corresponded to changes from baseline to Day 28 for each spirometry measure using a prespecified analysis of covariance (ANCOVA) model that included effects for center, treatment, baseline spirometry value, age, sex, and baseline Pseudomonas aeruginosa status. Baseline spirometry values corresponded to the average of the two values obtained before and closest to randomization, for variance reduction. Patients who discontinued early had their last post-baseline spirometry assessment carried forward for the primary, intent-to-treat analyses (Study Endpoint). A repeated measurements analysis was performed using mixed models, where treatment effects averaging across days in the study and the treatment by time interaction were assessed. Descriptive statistics for percent change, defined as change from baseline to Day 28 divided by baseline, were also generated. Tests of treatment effects with respect to improvement in each HRCT parameter (averaged across lobes and readers) at Day 28 were performed using ANCOVA models controlling for center, age, and sex. Cochran-Mantel-Haenszel statistics, using row mean scores and stratified by center, were generated to evaluate treatment effects with respect to each respiratory symptom. The association between treatment and proportion of patients experiencing a pulmonary exacerbation was tested via Cochran-Mantel-Haenszel chi-squared tests stratified by center. Statistical testing was not performed for change in sputum weight due to the small number of patients (< 50%) expectorating any amount of sputum during the study. The target sample size of 80 patients was based on safety objectives but would provide approximately 80% power to detect a difference of 0.20 L FEV1 between two treatment groups assuming an SD of 0.22 L. All statistical tests are two-sided, and nominal P values are presented without adjustment. The Bonferroni-Holm method was applied post hoc to account for multiplicity due to both multiple dose comparisons and multiple endpoints (FEV1, FVC, FEV1/FVC, and FEF25%-75%) (29). For all other comparisons, a significance level of 0.05 was used. Statistical analyses were performed using SAS Version 8.2 (SAS Institute, Cary NC). Additional details are provided in the online supplement.
One hundred seven patients were screened at 14 centers between April 11, 2003 and December 30, 2003. Eighty-nine patients met the eligibility criteria and received at least one dose of study drug (Table 1), and all but five patients (94%) completed the study. Completion rates and compliance with thrice-daily dosing were comparable among treatment groups.
Patients were primarily white, with a median age of 14.0 (range 8–45) years and a mean FEV1 of 93.0 (± 11.07) percent of predicted value (Table 2). Sixty-two (70%) patients were less than 18 years old. There appeared to be some imbalance with respect to sex across treatment groups, with the placebo and 20 mg group having more males (P = 0.067). A trend toward imbalance was also observed with respect to patients hospitalized within the previous year for an exacerbation (P = 0.054), with patients in the 40 mg group having a somewhat higher rate and patients in the 60 mg group having a somewhat lower rate than patients in the other treatment groups.
Safety Denufosol was well tolerated, and no dose–response trends were observed with respect to the incidence of adverse events or other safety parameters (Table 3). The most frequently occurring adverse event was cough. Five patients (two on placebo, two on 40 mg, and one on 60 mg) discontinued the study early due to adverse events, including hemoptysis, pulmonary function test decrease, lung infiltration, and cough. Two serious adverse events were reported in randomized patients: diagnosis of Hodgkin's lymphoma during screening (60 mg patient) and pulmonary exacerbation (40 mg patient); neither was considered related to treatment.
At 2 hours after dose on Day 1, there was an acute, transient decline in lung function relative to pre-dose levels among some patients receiving denufosol (Figure 2). The 40 mg and 60 mg groups had a significant relative decrease in FEV1 compared with placebo. Significant differences between placebo and 60 mg were also seen for FEF25%-75% and FEV1/FVC at 2 hours after dose. By 5 hours after dose, differences among treatment groups in lung function declines were no longer statistically significant. There was a statistically significant but clinically negligible change from baseline in oxyhemoglobin saturation at 60 minutes after dose on Day 14 in the 40 mg dose group compared with placebo (change from baseline = –0.2 ± 1.29 percent on room air for denufosol 40 mg versus 0.6 ± 1.19% on room air for placebo; P = 0.048). Otherwise, oxyhemoglobin saturation values were not significantly different between placebo and any denufosol dose group at any visit. Ten, 26, 23, and 30% of patients receiving placebo, 20, 40, and 60 mg, respectively, used a rescue bronchodilator at any time during the study (P = 0.245, 0.412, and 0.137 for 20, 40, and 60 mg compared with placebo, respectively). There was no pattern suggestive of an adverse drug effect with respect to clinical laboratory parameters, vital signs, or physical exam.
Efficacy Lung function. The results of the primary efficacy analysis demonstrated that denufosol patients had significantly greater changes in lung function at Study Endpoint, as measured by FEV1, FEF25%-75%, FVC, and FEV1/FVC, than patients treated with placebo (Table 4; pooled active doses versus placebo). The change from baseline to end of study in FEV1 was significantly greater for the 20 mg and 60 mg dose groups as compared with placebo (0.18 ± 0.059 L and 0.15 ± 0.062 L, respectively). Statistical significance was not reached in the 40 mg dose group, but the difference in adjusted means indicated a trend toward improvement (0.09 ± 0.060 L). The significance of these findings is maintained after taking into account multiple comparisons. In terms of FEV1 percent of predicted value, the average difference (± SE) from placebo in the change from baseline to Day 28 was 5.9 (± 2.21)% predicted for the 20 mg dose group (P = 0.010), 2.5 (± 2.29)% predicted for 40 mg (P = 0.281), and 4.2 (± 2.27)% predicted for 60 mg (P = 0.066), adjusting for site and baseline percent predicted value via ANCOVA.
Because the number of early discontinuations in the placebo group were unexpected, post hoc analyses of FEV1 were conducted to examine the robustness of the positive findings from the primary efficacy analysis. The results for only those patients completing the Day 28 visit showed smaller treatment effects and less statistical significance, due to the fact that the placebo dropouts were excluded. Differences from placebo in change from baseline FEV1 were 0.13 (± 0.053) L, 0.05 (± 0.055) L, and 0.11 (± 0.055) L for the 20, 40, and 60 mg groups, respectively, with corresponding P values of 0.019, 0.322, and 0.047 among those completing 28 d of treatment. Results similar to those observed for FEV1 were also seen for the other three spirometry measures (Table 4). In particular, treatment effects with respect to FEF25%-75%, a measure of flow at low lung volumes, thought a priori to be a good indicator of denufosol's pharmacologic activity, showed significant improvement for patients receiving 20 mg denufosol compared with placebo (difference of 0.40 ± 0.132 L/s in change from baseline), and trends toward significance for those receiving the 60 mg dose (0.30 ± 0.139 L/s in change from baseline) after adjusting for multiple dose comparisons (details provided in the online supplement). Figure 3 depicts the mean percent change from baseline in FEV1 for each week in the study. As can be seen from these weekly trends, little separation is observed between placebo and any active dose of denufosol until Day 21. Among those completing 28 days of treatment, mean change from baseline (± SD) in FEV1 was –2.6 (± 6.29)% for placebo and 1.2 (± 7.75)%, –1.0 (± 6.26)%, and 0.4 (± 5.43)% for 20, 40, and 60 mg, respectively. Only 33% of placebo patients experienced a positive improvement in FEV1 at Study Endpoint relative to baseline value, while over 50% of the patients in the 20 mg group showed positive improvement. Repeated measures analysis of FEV1 yielded positive results, with denufosol-treated patients experiencing significantly greater improvement compared with placebo, averaging across days in the study (P = 0.020). A test of differences in linear slopes across treatment groups was not significant (P = 0.832).
It was hypothesized a priori that analyzing the average of the values from the three best spirometry maneuvers would yield reduced variance and increased power. This was indeed the case, as indicated by the significance level of the global test of any active dose versus placebo for FEV1 (P = 0.003). Treatment effects for each active dose were increased somewhat by averaging across the three best spirometric maneuvers; for example, the difference in FEV1 for 20 mg patients versus placebo was 0.20 ± 0.062 liters (P = 0.002) with this analysis. No evidence of differential treatment effects across study centers was seen. The treatment by center interaction for FEV1 was not significant (P = 0.9872). Treatment by subgroup interactions are described in the online supplement.
Symptoms reported on the CF questionnaire.
HRCT.
Pulmonary exacerbations.
Initial events in the pathophysiology of CF lung disease reflect abnormal ion transport and defective mucus clearance, which results in chronic bacterial infection of the airways. There have been therapeutic advances that treat the secondary effects of infection and inflammation in CF airways, but P2Y2 agonists offer the potential to treat "primary" pathobiological defects. Specifically, aerosolized P2Y2 agonists target defective mucus clearance in CF through multiple mechanisms to (1) stimulate chloride (and liquid) secretion via non-CFTR mechanisms, (2) inhibit sodium (and liquid) absorption, (3) induce mucin secretion from surface goblet cells, and (4) stimulate ciliary beat frequency. The primary objective of this study was to test the safety of repeated (three times daily) aerosolized administration of denufosol over 28 days, with secondary objectives to assess biological effects of denufosol, compared with placebo. The study design was developed in conjunction with the CFF-sponsored TDN, and included not only traditional measures of safety and efficacy (e.g., spirometry), but also explored methods to assess mucus clearance (retention), using HRCT of the chest. The effects of P2Y2 agonists are predicted to improve mucus clearance and lung function with chronic treatment. Thus, this relatively short study focused on patients with mild lung disease who were not taking inhaled antibiotics to maximize sensitivity to detect any biological effect of denufosol on spirometry. This study shows that inhaled (three times daily) denufosol is safe and tolerable over 28 days at doses up to 60 mg in adult and pediatric patients with mild CF lung disease. The most common adverse events were cough, headache, pharyngitis, and nasal congestion, with incidences similar among all three denufosol dose groups and placebo. Self-reports of cough at night were more common among patients at the 60 mg dose, but, this trend was present at baseline. There were acute, but transient, decreases in FEV1, which is consistent with the mechanism of action for this P2Y2 agonist. Discontinuations due to adverse events were similar for the placebo and denufosol groups, and the two serious adverse events were not related to treatment. This study also provides evidence of biological efficacy on measures of airflow obstruction, in spite of the relatively short duration of treatment. Statistically significant findings for all four spirometric parameters, in comparisons of pooled doses versus placebo, were seen (Table 4 and Figure 3), even after post hoc adjustments for testing multiple spirometric endpoints and multiple doses using the Bonferroni-Holm method.
In terms of relative improvement (i.e., percentage change from baseline for denufosol- versus placebo-treated groups), only modest treatment differences were noted. For example, with respect to FEV1, there was an improvement of 1.2% from baseline for 20 mg denufosol patients, versus a 2.6% decline among placebo patients, for a difference from placebo in relative improvement of 3.8% among those completing Day 28. The small levels of relative improvement may be due in part to the mild lung impairment in this study population at baseline. The magnitude of denufosol treatment effects relative to placebo are similar to those seen with azithromycin ( The decline of FEV1 among placebo patients over 28 days is most likely due to the phenomenon of regression to the mean. Because entry criteria required patients to meet or exceed an FEV1 of 75% predicted at screening, subsequent measurements of FEV1 would be expected to be lower due to within-patient variability over time. A similar phenomenon occurred in PEIT, where patients with baseline FEV1 of at least 85% predicted were enrolled. During the first 12 weeks of the PEIT study, the FEV1 of placebo patients declined by over 2% predicted, as would be anticipated with a high bar for recruitment. In the absence of baseline imbalances with respect to strong predictors of treatment response, and assuming a valid randomization algorithm, differences after randomization in a clinical trial can be attributed to the treatments administered. That is, a decline similar to that observed for placebo-treated patients would be expected among denufosol-treated patients in the absence of treatment effects, based on the between-groups analysis. The absence of a linear dose–response relationship for tolerability and efficacy (FEV1) is perhaps not surprising, when pre-clinical data are considered. All three doses were expected to achieve an initial concentration of drug on airway surfaces in excess of the maximally effective concentration, based on dose effects on nasal potential difference (Cl– conductance) in patients with CF and in normal subjects (33). The results of this phase 2 study are consistent with a priori hypotheses based on the nasal potential difference results, namely that all three active doses would be different from placebo, with little difference among the three doses. The rationale for dose selection was appropriate for this stage of clinical development of denufosol, where a key objective is to detect a signal for efficacy that would justify further study of the compound. The HRCT scores did not provide evidence of efficacy of denufosol. The use of HRCT as an early indicator of efficacy in CF is an area of active investigation. Two placebo-controlled studies have shown improvement after therapy, both with dornase alfa (34, 35). The treatment period in both studies was 3 months or greater. Air trapping decreased in one study conducted in a single center with spirometer control. Both factors would be expected to decrease variability when compared with this multicenter trial in which CT scanning was performed without lung volume control. Mucus plugging, a finding that would be expected to be less sensitive to technique, did not change in this study or in the prior studies. In this study, mucus plugging was only identified in 20% of lobes, which may have limited the sensitivity of this indicator.
We expect longer term studies of denufosol to show reductions in other endpoints, such as pulmonary exacerbations. A large clinical trial of inhaled hypertonic saline in CF demonstrated only modest improvements in FEV1 ( In conclusion, aerosolized denufosol given three times daily in doses up to 60 mg was well tolerated over 28 days in both adult and pediatric patients with mild lung disease. Importantly, there was evidence of a biological effect on multiple measures of airflow mechanics in comparison to placebo. The use of P2Y2 agonists to improve mechanisms of mucus clearance in CF holds promise for targeting the primary "root cause" of the disease, with the ultimate goal of initiating therapy very early in life to prevent or slow the initial pathobiology of CF airways disease.
The authors acknowledge the significant contributions of the Cystic Fibrosis Foundation Therapeutics, Inc., in particular Dr. Bob Beall and Dr. Preston Campbell based in Bethesda, Maryland as financial co-sponsors of this study; the Cystic Fibrosis Therapeutics Development Network and Children's Hospital & Regional Medical Center based in Seattle, Washington as co-managers of the clinical operations for this study; Dr. Michael Knowles and Dr. Richard Boucher at the University of North Carolina at Chapel Hill for their assistance in the planning and implementation of this study; Dr. Anthony Fox with the EDB Group in Carlsbad, California for his work as the Medical Consultant to Inspire Pharmaceuticals, Inc.; Dr. Scott Davis, Dr. David Orenstein, Dr. Lisa Saiman, and Michael Kosorok, Ph.D. as members of the Data Monitoring Committee; and the 14 cystic fibrosis centers across the USA that participated in the study (see The Working Group). Members of the Working Group were as follows: Dr. Robin Deterding at The Children's Hospital, Denver, CO and the lead Principal Investigator (GCRC grant number M01-RR00069); Dr. Cori Daines at Children's Hospital Medical Center, Cincinnati, OH (GCRC grant number M01–08084); Dr. Hank Dorkin at Massachusetts General Hospital, Boston, MA; Dr. Ronald Gibson at the University of Washington Medical Center, and Children's Hospital and Regional Medical Center, Seattle, WA (GCRC grant number M01-EE-00037); Dr. Christopher Harris at Vanderbilt University, Nashville, TN; Dr. Noreen Henig at Leland Stanford Junior University, Palo Alto, CA (GCRC grant number 5M01-RR00070); Dr. Peter Hiatt at Baylor College of Medicine, Houston, TX (GCRC grant number Texas Children's: M01RR00188, Baylor 13081); Dr. Susanna McColley at Children's Memorial Hospital, Chicago, IL, Dr. Karen McCoy at Children's Research Institute, Columbus, OH, Dr. Carlos Milla at the University of Minnesota, Minneapolis, MN (GCRC grant number MO1-RR00400); Dr. George Retsch-Bogart at the University of North Carolina at Chapel Hill, Chapel Hill, NC (GCRC grant number 5–32589); Dr. David Waltz at Children's Hospital Boston, Boston, MA (GCRC grant number MO1-RR02172); Dr. Robert Wilmott at St. Louis University, St. Louis, MO, Dr. Pamela L Zeitlin at The Johns Hopkins University, Baltimore, MD (GCRC grant number MO1-RR000052).
* A complete list of members of the Inspire 08–103 Working Group may be found before the beginning of the REFERENCES.  This research was supported by Inspire Pharmaceuticals, Inc. and The Cystic Fibrosis Foundation Therapeutics, Inc. This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org. Originally Published in Press as DOI: 10.1164/rccm.200608-1238OC on April 19, 2007 Conflict of Interest Statement: R.R.D. received $9,818.26 from Inspire for consulting and presenting at training meetings and conferences from 2002–2006. L.L. was an employee of Inspire Pharmaceuticals from September 2001 through December 2006 and is currently a consultant to Inspire. J.M.E. has been a part-time employee of Inspire Pharmaceuticals, Inc. since May 2004 and has stock options through Inspire. D.W.M. is an employee of Inspire Pharmaceuticals, Inc. since April 2000 and has stock options through Inspire. S.J.C. was hired as a consultant in February 2006 to perform monitoring activities and employed by the Therapeutics Development Network Coordinating Center (TDN-CC), from March 2000 through June 2004, which was contracted by Inspire Pharmaceuticals to manage and monitor clinical trials. A.S.B. received $5,690.59 from Inspire Pharmaceuticals from 8/6/2002–12/16/2005 for consulting, speaking at the study training meeting, and for expenses for presenting results at the North American Cystic Fibrosis Conference. S.P.M. was employed by TDNCC from June 2002–December 2005. The TDNCC was contracted by Inspire to manage and monitor the 08–103 study, as well as provide statistical consulting. B.W.R. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Received in original form August 30, 2006; accepted in final form April 19, 2007
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
AT A GLANCE COMMENTARY
75% of predicted normal value and not treated with inhaled antibiotics for the past 30 days were randomized to receive one of three doses of denufosol or placebo administered three times daily. 
4%) at 4 weeks and those observed in the Pulmozyme (dornase alfa) Early Intervention Trial (PEIT; 
F508 CFTR-expressing human bronchial epithelia. Am J Physiol 1999;276:827–837.
