Effectiveness of a computerized decision support system for ...

Hemodialysis International 2016; 00:00–00

Effectiveness of a computerized decision support system for anticoagulation management in hemodialysis patients: A before–after study Edward G. CLARK,1 Marc A. RODGER,2 Tim O. RAMSAY,3 Greg A. KNOLL1 1

Division of Nephrology, Department of Medicine and Kidney Research Centre, Ottawa Hospital Research Institute, University of Ottawa, Ottawa, Ontario, Canada; 2Division of Hematology, Department of Medicine, and Ottawa Hospital Research Institute, University of Ottawa, Ottawa, Ontario, Canada; 3Clinical Epidemiology Program, Ottawa Hospital Research Institute, University of Ottawa, Ottawa, Ontario, Canada

Abstract Introduction The risk-benefit profile for warfarin anticoagulation in hemodialysis (HD) patients differs compared with the non-HD population. HD patients are at increased risk of both thromboembolism and bleeding related to anticoagulation therapy. In addition, anticoagulation control may be more difficult to achieve in the HD population due to frequent comorbidities, subclinical Vitamin K deficiency, altered pharmacokinetics due to uremia and the concurrent use of multiple medications. While computerized decision support systems (CDSS) to assist with anticoagulation management are safe and effective in the non-HD population, they have not been well studied in HD outpatients. Methods A before–after study compared anticoagulation control for HD outpatients receiving warfarin at a tertiary medical center in Canada during two time periods: an initial period of nephrologist-led anticoagulation management and a second period after implementation of a pharmacist-led, CDSS-assisted anticoagulation management strategy. Findings Forty-two patients were included. Following implementation of the CDSS-assisted strategy, there was no significant change in median therapeutic time-in-range (3.7% difference (IQR, 29.5% to 20.6%); P 5 0.247). Median change in INR testing frequency was 1.2 (IQR, 0.1–2.2; P 5 0.0001) fewer tests per patient per month, which equates to approximately 15 fewer tests per patient per year. Adverse events were similar. Discussion Implementing a CDSS-assisted strategy for anticoagulation management in HD outpatients is effective. Doing so may lead to modest cost savings related to less frequent INR testing. Key words: Warfarin, hemodialysis, computer-assisted decision making, computer-assisted drug therapy, drug monitoring

INTRODUCTION Correspondence to: Edward Clark, MD, MSc, FRCPC, Division of Nephrology - The Ottawa Hospital, Riverside Campus, 1967 Riverside Drive, Ottawa, ON, Canada K1H 7W9. E-mail: [email protected] Conflict of Interest: There is no conflict of interest. Disclosure of grants or other funding: The Ottawa Hospital Department of Medicine Research Priority Grant Program.

To reduce the risk of bleeding due to overanticoagulation and the risk of thromboembolic complications with under-anticoagulation,1 the therapeutic use of warfarin necessitates careful dose-adjustment and monitoring. In patients with end-stage-renal-disease on hemodialysis (HD) the risk of venous thromboembolism

C 2016 The Authors. Hemodialysis International published by Wiley Periodicals, Inc. on behalf of International Society for V

Hemodialysis. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. DOI:10.1111/hdi.12411 1

Clark et al.

(VTE)2,3 and atrial fibrillation4–6 is increased relative to the general population. At the same time, therapeutic anticoagulation in HD patients may be more likely to be associated with bleeding complications.7 There are multiple reasons why anticoagulation control may be more difficult to achieve in the HD population including frequent comorbidities, nutritional deficiencies (including subclinical Vitamin K deficiency),8 altered pharmacokinetics due to uremia and the concurrent use of multiple medications. Computerized decision support systems (CDSSs) have been shown to be effective tools to assist in the management of warfarin dosing and INR testing.9 Given that the HD population may have a different risk and benefit profile for therapeutic warfarin anticoagulation than the general population,10 the use of a CDSS for managing warfarin anticoagulation in HD patients has the potential for more or less effective anticoagulation control than has been observed in studies of less-selected populations. On this basis, we sought to determine the impact of CDSSguided warfarin management specifically for HD patients whose anticoagulation was being managed by nephrologists.

MATERIALS AND METHODS Study design This study was a quasiexperimental, before–after study with prospective and retrospective data-collection.

Objectives The primary objective was to determine if implementation of a CDSS to manage systemic warfarin anticoagulation in HD patients was associated with a change in the percentage of time that INR is in therapeutic range compared with nephrologist-led anticoagulation management. Secondary objectives were to compare the frequency with which laboratory testing of the INR was undertaken before and after implementation of a CDSS. Lastly, we sought to compare the occurrence of adverse events related to poor anticoagulation control before and after implementation. A formal safety analysis was not undertaken due to the likelihood of insufficient statistical power.

The intervention The intervention was the implementation of a pharmacistled, CDSS (DawnAC, 4S Information Systems Ltd., 2

Milnthorpe, England) assisted warfarin anticoagulation management strategy for HD patients. “Usual care” was adjustment of warfarin dosing and INR testing according to the prior existing strategy (i.e., nephrologists adjusting warfarin and INR testing as per their own practice without a specific protocol). For CDSS-based management (the intervention), suggestions of the CDSS regarding changes in warfarin dose and/or timing of next INR testing were relayed by the pharmacist responsible for operating the CDSS to the HD-unit where the patient usually received dialysis and via telephone contact with the patient to be implemented without direct physician involvement.

Data collection Data collection took place between May 5th, 2011 and June 18th, 2014. For individual patients, any INR data from the 3 months prior to them having their anticoagulation management switched from “usual care” to the CDSS was collected from the hospital wide electronic health record. This 3 month period was defined as the preintervention period. In addition, a retrospective review (including review of a separate specialized clinical database for dialysis patients) was conducted to obtain demographic and other baseline information. From the date following implementation of CDSS management of their anticoagulation, INR data were collected for 9 months, defined as the postintervention period. Throughout the follow-up period, the clinical databases/charts were reviewed on a monthly basis to determine if events, as defined below, had occurred with additional information being obtained from dialysis unit staff if necessary.

Recruitment All eligible patients were identified by monthly screening for active prescriptions for warfarin entered in the dialysis patient database (which keeps up-to-date records on medications for all HD patients). There were approximately 650 dialysis patients at our Center during the study period. The study was approved by the Ottawa Hospital Research Ethics Board. The requirement for consent was waived as all data abstracted from patient records was anonymized and deidentified prior to analysis.

Inclusion and exclusion criteria Patients were included in the study if they were 18 years of age or older; prevalent or incident chronic HD patients receiving regular HD at one of the dialysis units affiliated Hemodialysis International 2016; 00:00–00

Computerized anticoagulation management for hemodialysis patients

with The Ottawa Hospital; and who were being prescribed warfarin for any indication. Patients were excluded if they were receiving oral anticoagulation other than warfarin; their warfarin anticoagulation was being managed by a physician other than the nephrologist(s) covering their HD unit (e.g., family physician); INR testing data was not available from both preintervention and postintervention periods (i.e., before and after initiation of CDSS-based anticoagulation management).

Primary outcome measure The primary outcome measure was ‘Time-in-Range’ (TIR). A multistep method was used to determine the TIR. As described by Rosendaal et al.,11 linear interpolation was used to calculate the INR values for the days between actual measurements. For each patient, the proportion of days, during the specified observation period, in which the INR was within the therapeutic range, was calculated. This value, multiplied by 100%, resulted in the TIR.11

Secondary outcome measures Secondary outcomes included the frequency of INR testing: the number of times an INR test was conducted during the specified observation period divided by the number of days in that observation period. The percentage of INR tests in the therapeutic range was also assessed. Other secondary outcomes were related to bleeding and thrombotic events. Major bleeding was defined according to definitions of the International Society on Thrombosis and Haemostasis:12 associated with a fall in hemoglobin of 20 g/L or more, or; leading to a transfusion of two or more units of packed red blood cells or whole blood, or; occurring in a critical site (intracranial, intraspinal, intraocular, pericardial, intraarticular, intramuscular with compartment syndrome, retroperitoneal), or; contributing to death. VTE events included: clinical or radiologic diagnosis of deep vein thrombosis; clinical or radiologic diagnosis of pulmonary embolus; cerebrovascular events (including clinical diagnosis of a transient ischemic attack, clinical diagnosis of a reversible ischemic neurologic deficit and clinical or radiologic diagnosis of a cerebrovascular accident (CVA)) and; unexpected death attributed to VTE, major bleeding or CVA.

to compare the preintervention and postintervention frequency of INR testing. The reason for using this nonparametric test was that TIR (the primary outcome measure) is a proportion and is, therefore, not a continuous variable with a normal distribution. Furthermore, TIRs (preintervention and postimplementation) were not anticipated to have a near-normal distribution within the confines of the TIR range of 0% to 100%. Subgroup analyses (such as, according to target INR) were not performed due to the small numbers of patients within subgroups. Given the likelihood that TIR in the initial postintervention phase would be influenced by anti-coagulation control in the preintervention phase, the initial 14 days postimplementation of the CDSS was censored from observation. Time periods during which patients were admitted to hospital were censored from observation in addition to 14 days following hospitalizations (not including emergency department visits). Time periods where warfarin was temporarily held in anticipation of a procedure were censored from observation in addition to 14 days following resumption of warfarin. In addition to performing a paired comparison of anticoagulation parameters during the preintervention and postintervention periods for all included patients (as described above), adverse events were tracked and reported for all patients who were initially included in the study but were later excluded for not having data available from the postintervention period (i.e., due to death or discontinuation of warfarin prior to the intervention). Statistical analysis was undertaken using SAS 9.4 (SAS Institute Inc.; Cary, NC, USA).

Power calculations With respect to the primary outcome and the frequency of INR testing, post hoc power calculations were conducted for paired t-tests assessing mean differences preintervention vs. postintervention using a two-tailed alpha of 0.05. We had previously determined that the number of subjects (with each subject contributing both preintervention and postintervention TIR values) required to have 80% power to detect a 10% difference in mean TIR with 95% confidence was 41.

Statistical analysis

RESULTS

The TIR for the 2 study periods (preintervention and postintervention) was compared using a Wilcoxon signed-rank test for paired samples assuming a two-sided alpha of 0.05. A Wilcoxon signed-rank test was also used

Patient characteristics and follow-up


Over the study period, 46 patients were identified for inclusion into the study on the basis of an active Warfarin 3

Clark et al.

Table 1 Baseline patient characteristics Characteristic n 5 42 Female (%) 19 (45.2) Median age in years 70 (62–77) (Interquartile range) Primary indication for anticoagulation [target INR] (%) Atrial fibrillation [2–3] 23 (54.8) Venous thromboembolism [2–3] 11 (26.2) Mechanical heart valve [2.5–3.5] 4 (9.5) Bioprosthetic heart valve [2–3] 2 (4.8) Maintenance of dialysis 2 (4.8) catheter patency [1.5–.5] Duration of prior anticoagulation in years (%) 5 4 (9.5) Duration of prior hemodialysis in years (%) 5 13 (30.9) Primary cause of end-stage-renal-disease (%) Diabetic nephropathy 21 (50) Ischemic nephropathy 9 (21.4) Glomerulonephritis 6 (14.3) Hereditary nephropathy 2 (4.8) Obstructive uropathy 1 (2.4) Unknown 3 (7.1) Anti-platelet medication use (%) Acetylsalicylic acid (ASA) 14 (35.7) Clopidogrel 3 (7.1) SD 5 standard deviation.

prescription. All identified patients met the inclusion criteria and were entered into the study. None of the identified patients initially met any exclusion criteria. Four patients were subsequently excluded for not having postimplementation INR data available for the following reasons: warfarin stopped due to upper gastrointestinal bleed with INR 4.3 (n 5 1); death due to septic and cardiogenic shock (n 5 1); death presumed due to sudden cardiac death at home (most recent INR had been in the therapeutic range) (n 5 1); and death following admission with a myocardial infarction (n 5 1). Table 1 reports baseline characteristics of the 42 patients included for complete data analysis. Approximately half of the patients were female. The median age (IQR) was 70 years (62–77). Thirty-nine patients (92.9%) were white, 1 (2.4%) was aboriginal, 1 (2.4%) was black and 1 (2.4%) was South Asian. Atrial fibrillation was the most common indication for anticoagulation (59.5%). All patients except one were being treated with heparin while on HD at the time of enrollment. 4

Table 2 summarizes the duration of patient observation included for analysis after censoring follow-up time immediately after implementation of the intervention, for procedures and during hospitalizations. Total included observation time was 32.8 years out of a total of 42 years of follow-up (78.1%). No patients were lost to follow-up.

Time-in-range The median preintervention TIR was 56.5% (IQR, 44.1– 73.6%) and the median postintervention TIR was 60.9% (IQR, 47.1–75.6%). The median of the differences in TIR was a 3.7% (IQR, 29.5% to 20.6%; P 5 0.247) postintervention increase. Figure 1 shows preintervention and postintervention TIRs in ascending order (not linked to same patient).

Frequency of INR tests-in-range The median percentage of INR tests in the therapeutic range was 49.1% (IQR, 33.8–71.4%) in the preintervention period. In the postintervention period, the median percentage of INR tests in the therapeutic range increased to 50.5% (IQR, 40.9–64.45%; P 5 0.689).

Frequency of INR testing The preintervention median number of INR tests was 5.0 (IQR, 4.3 to 6.3) tests per patient per month and the postintervention median was 3.8 (IQR, 2.8 to 5.1) tests per patient per month. The median of the differences was 1.2 (IQR, 0.1 to 2.2; P 5 0.0001) fewer tests per patient per month during the postintervention period. This equates to approximately 15 fewer INR tests per patient per year.

Statistical power With 42 subjects contributing preintervention and postintervention data, this study had 80% power to detect a Table 2 Summary of duration of patient observation Follow-up

Duration

Mean number of preintervention months included for observation (SD) Mean number of postintervention months included for observation (SD) Total preintervention observation in months Total postintervention observation in months Total observation time in months

2.5 (0.84) 6.9 (2.7) 103.1 290.6 393.7



Figure 1 Preintervention and postintervention time-in-range values (n 5 42). For each time period, values are arranged in ascending order and are not paired according to individual patients.

9.6% difference in preintervention vs. postintervention TIR. As well, it had 80% power to detect a 2.8% difference in preintervention vs. postintervention frequency of INR testing.

postintervention period or, 0.12 and 0.21 events per year of follow-up, respectively. Table 4 details the major bleeding events.

Adverse events

Table 4 Major bleeding events

Adverse events were reported for the 42 patients included for complete data analysis as well as for those patients who were initially enrolled but subsequently excluded since they did not have TIR data from the postintervention period. There were 3 deaths in the preintervention period and 4 in the postintervention period. All deaths are detailed in Table 3. None were attributed to pulmonary embolism or major bleeding. There was 1 major bleeding event in the preintervention period and 5 in the Table 3 Deaths Study phase Preintervention

Postintervention

Cause of death

n

Septic and cardiogenic shock Following admission with NSTEMI Presumed sudden cardiac death at home Sudden cardiac death Septic shock secondary to pneumonia Palliative withdrawl from dialysis treatment

1


1 1 1 1 2

Preintervention period Admitted with bleeding diathesis manifested as gingival bleeding and extensive bruising. Requiring transfusion of 2U PRBCs and reversal of INR. No obvious precipitant for increased INR beyond warfarin prescription. Postintervention period Admitted with right renal hemorrhage following spontaneous rupture of preexisting angiomyolipoma. Seen in ED with epistaxis. Received transfusion 2of U PRBCs. Admitted with lower gastrointestinal bleed presumed secondary to angiodysplasia. Seen urgently in ophthalmology clinic for decreased vision in right eye found to be secondary to intraocular bleeding. Admitted for epistaxis in context of thrombocytopenia due to idiopathic thrombocytopenic purpura.

INRa 7.1

2.1

3.5 2.9 2.0

1.0

a

Most recent INR preceding the event. ED 5 emergency department; U 5 units; PRBCs 5 packed red blood cells; INR 5 international normalized ratio.

5

Clark et al.

DISCUSSION A change from a nephrologist-led anticoagulation management strategy to one involving the use of a pharmacist-led CDSS resulted in no significant change in anticoagulation control amongst hemodialysis outpatients on Warfarin; however, the CDSS-based strategy resulted in a significant reduction in the frequency of INR testing. The risk of bleeding and thrombotic events in the preintervention and postintervention periods was similar when taking into account longer follow-up time for the postintervention period compared with the preintervention period (9 months vs. 3 months). The use of CDSS-assisted maintenance warfarin dosing has been extensively evaluated in non-HD patients. Pooled analysis of 7 studies, which included 14,213 patients in total, showed a small but significant improvement in TIR with the use of a CDSS compared with not using one: the mean TIR improvement was 4.5% (95%CI, 2.4–6.7%).13 This is similar to the difference in median TIR of 3.7% observed in our study. While the TIR is an extrapolated value, it is notable that the percentage of tests-in-range was, similarly, not significantly different during the preintervention vs. postintervention periods. Overall, the results are also consistent with those of much larger studies that have reported CDSS-managed maintenance warfarin dosing does not have a significant impact on the risks of VTE, major bleeding and death for the non-HD population of patients on warfarin anticoagulation.13 This study was not designed to, or powered to, detect a difference in clinical endpoints. In addition, it was not adequately powered to detect what might be clinically meaningful differences in TIR for this particular patient population. Past studies have suggested that achieving a TIR of 65% or more is associated with a reduced risk of complications.1,14 The median postintervention TIR observed in this study is lower than the average of 64% reported for non-HD patients managed using a nurse- or pharmacist-led anticoagulation strategy.15 Nonetheless, it should be noted that the postintervention TIR was not significantly worse than 65% (mean postintervention TIR of 60.6% with a 95% CI of 1/25.7% (not shown)). A similar study evaluating the use of an electronic nomogram to manage warfarin anticoagulation in HD-patients was reported by Thomson et al. in 2011.16 This before– after study compared nephrologist-led anticoagulation management to a nurse-led, electronic nomogram-guided strategy. In general, the results of our study were similar. Thomson et al. showed that the use of the nomogram did not result in a significant change in TIR but did significantly reduce INR testing.16 6

While a formal assessment of cost effectiveness was beyond the scope of this study, there are potential cost implications to less frequent INR testing. In Ontario, the cost of a single outpatient INR test consists of a C$6.20 laboratory fee plus an additional C$7.76 in documentation and administration costs, for a total cost of C$13.96 per test.17 As such, reducing testing in HD patients on Warfarin by 15 tests per year would result in a cost savings of C$209.40 per patient on warfarin, which is likely to be insignificant relative to the overall high cost associated with HD-treatment on an annual basis. Nonetheless, additional cost benefits may include reduced physician and nursing time dedicated to anticoagulation management due to less frequent testing and less direct involvement with the pharmacist-led, CDSS-assisted strategy. Assuming that clinical outcomes are equivalent, as shown by our study, Thomson et al.16 and in much larger studies in the non-HD population,13 potential cost benefits are balanced by the fact that the CDSS software is expensive:13 the largest cost-effectiveness study of CDSSs (that included patients managed using the DAWN AC CDSS which we also used) concluded that its use was only costeffective through a reduction of individual patient costs on a large scale.18 In the case of our study, at a center where a pharmacist-led CDSS-assisted strategy was already being employed for a larger non-HD population of patients on warfarin, cost-savings justify ongoing use of this strategy for the HD-population. More generally, our findings suggest that implementation of CDSS-assisted management of anticoagulation for HD patients can be justified if done in conjunction with a program managing the anticoagulation of a larger non-HD population rather than implementing it as a stand-alone program for a relatively small number of patients. This study has many important limitations including being underpowered to detect small but potentially clinically meaningful differences in TIR. It is also possible but unlikely that nephrologists and nurses involved in managing HD-patients’ anticoagulation management altered their usual practices in such a way as to bias patients’ preintervention TIRs (i.e., a Hawthorne effect). The same can be stated about the possibility of a Hawthorne effect influencing the performance of the pharmacists involved in managing anticoagulation during the postintervention period. Given that data collection was performed using on-line databases and this study did not have an obvious physical presence or other reminders that it was being conducted, either in HD-units or in the offices where the pharmacists managing the CDSS were based, any Hawthorne effect is unlikely. Hemodialysis International 2016; 00:00–00


In conclusion, we have shown that a CDSS-assisted strategy was not associated with a change in warfarin anticoagulation control but was associated with a significant decrease in the frequency of INR testing in HD patients compared with usual care by nephrologists. Given the likelihood of therapeutic efficacy, equivalent safety, and the potential for cost savings, implementation of a CDSSassisted anticoagulation strategy for HD patients should be considered at centers where a program of CDSS-assisted anticoagulation already exists for the non-HD population.

10

ACKNOWLEDGMENTS

11

The authors thank Jessica Wagner, Judy Cheeseman, Scott Mullen, Edita Delic and Hannah Trottier for dedicated assistance with data collection.

8

9

12

Manuscript received July 2015; revised January 2016.

REFERENCES 1 Oake N, Jennings A, Forster AJ, Fergusson D, Doucette S, van Walraven C. Anticoagulation intensity and outcomes among patients prescribed oral anticoagulant therapy: A systematic review and meta-analysis. CMAJ. 2008; 179:235–244. 2 Ocak G, van Stralen KJ, Rosendaal FR, et al. Mortality due to pulmonary embolism, myocardial infarction, and stroke among incident dialysis patients. J Thromb Haemost. 2012; 10:2484–2493. 3 Ocak G, Vossen CY, Rotmans JI, et al. Venous and arterial thrombosis in dialysis patients. Thromb Haemost. 2011;106: 1046–1052. 4 Atar I, Konas D, Acikel S, et al. Frequency of atrial fibrillation and factors related to its development in dialysis patients. Int J Cardiol. 2006; 106:47–51. 5 Genovesi S, Pogliani D, Faini A, et al. Prevalence of atrial fibrillation and associated factors in a population of long-term hemodialysis patients. Am J Kidney Dis. 2005; 46:897–902. 6 Winkelmayer WC, Patrick AR, Liu J, Brookhart MA, Setoguchi S. The increasing prevalence of atrial fibrillation among hemodialysis patients. J Am Soc Nephrol. 2011; 22:349–357. 7 Elliott MJ, Zimmerman D, Holden RM. Warfarin anticoagulation in hemodialysis patients: a systematic


13

14

15

16

17

18

review of bleeding rates. Am J kidney Dis. 2007; 50: 433–440. Pilkey RM, Morton AR, Boffa MB, et al. Subclinical vitamin K deficiency in hemodialysis patients. Am J Kidney Dis. 2007; 49:432–439. Ansell J, Hirsh J, Hylek E, Jacobson A, Crowther M, Palareti G. Pharmacology and management of the vitamin K antagonists: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines (8th Edition). Chest. 2008; 133:160S–198S. Bansal N. The debate on warfarin use in dialysis patients with atrial fibrillation: More fuel for the fire. Am J Kidney Dis. 2014; 64:677–680. Rosendaal FR, Cannegieter SC, van der Meer FJ, Briet E. A method to determine the optimal intensity of oral anticoagulant therapy. Thromb Haemost. 1993; 69:236– 239. Schulman S, Kearon C, Subcommittee on Control of Anticoagulation of the S, Standardization Committee of the International Society on T, Haemostasis. Definition of major bleeding in clinical investigations of antihemostatic medicinal products in non-surgical patients. J. Thromb. Haemost. 2005; 3:692–694. Holbrook A, Schulman S, Witt DM, et al. Evidencebased management of anticoagulant therapy: Antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians EvidenceBased Clinical Practice Guidelines. Chest. 2012; 141: e152S–184S. Oake N, Fergusson DA, Forster AJ, van Walraven C. Frequency of adverse events in patients with poor anticoagulation: A meta-analysis. CMAJ. 2007; 176:1589– 1594. van Walraven C, Jennings A, Oake N, Fergusson D, Forster AJ. Effect of study setting on anticoagulation control: A systematic review and metaregression. Chest. 2006; 129:1155–1166. Thomson BK, MacRae JM, Barnieh L, Zhang J, MacKay E, Manning MA, Hemmelgarn BR. Evaluation of an electronic warfarin nomogram for anticoagulation of hemodialysis patients. BMC Nephrol. 2011; 12:46. Ontario Ministry of Health and Longterm Care. Schedule of benefits for laboratory services. Available at: http://www. ontla.on.ca/library/repository/ser/303089/199901apr.pdf. Jowett S, Bryan S, Poller L, et al. The cost-effectiveness of computer-assisted anticoagulant dosage: Results from the European Action on Anticoagulation (EAA) multicentre study. J Thromb Haemost. 2009; 7:1482– 1490.

7