Oral surgery in anticoagulated patients without reducing the dose of oral anticoagulant: A prospective randomized study

J Oral Maxillofac 5427.32, 1996

Surg

Oral Surgery in An ticoagula ted Patients Without Reducing the Dose of OraI Anticoagulant: A Prospective J.C. SOUTO,*

Randomized

A. OLIVER,t I. ZUAZU-JAUSORO,* AND J. FONTCUBERTAS

Study A. VIVES,$

Purpose: This study assessed the risk associated with several schedules of perioperative treatment with coumadin in anticoagulated patients who underwent oral surgery. Patients and Methods: A prospective, randomized study compared bleeding complications with six perioperative schedules in 92 patients chronically treated with acenocoumarol. In three of the perioperative schedules, the dose was reduced before surgery and calcium heparin was added. In the other three, oral anticoagulation was not modified and heparin was not used. The groups also differed regarding the antifibrinolytic agents used and the postoperative measures applied. Results: Those schedules in which the oral anticoagulation was not modified preoperatively and an antifibrinolytic agent was applied locally both during and after surgery were not associated with a significantly higher odds ratio of bleeding complications than those in whom oral anticoagulation was reduced and calcium heparin was added preoperatively. Conclusions: In orally anticoagulated patients who undergo oral surgery, schedules that maintain the oral anticoagulant regimen and use local tranexamic acid as an antifibrinolytic agent postoperatively for 2 days are safe, simple, and less troublesome.

Patients with mechanical cardiac valve prosthesis or valvular heart disease have a high risk of thromboembolism. For untreated patients, the risk ranges from 2.5% to 10% per patient-year in those with a mechanical prosthesis and 1.5% to 4.7% per patient-year in those with mitral valve disease.’ Oral anticoagulant

agents (vitamin K antagonists) have been used to reduce the incidence of thromboembolic events.2-4 Bleeding is the main complication of these drugs. The risk of bleeding is influenced by the intensity of anticoagulant therapy.5 In patients undergoing general surgical procedures, no optimally designed studies evaluating the relationship between the risk of bleeding and the anticoagulant effect have been performed. However, the results of most studies suggest a relationship between bleeding and the anticoagulant effect.6 For this reason, before oral surgery is performed in patients taking oral anticoagulants, it is common practice to reduce or discontinue the anticoagulant medication to minimize the risk of bleeding. In an animal model, blood loss after dental extractions was significantly greater in animals anticoagulated at a therapeutic level than in non-anticoagulated control animals, but the study also showed that complete correction of

* Medical Staff, Unitat d’Hemostasia i Trombosi, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain. t Director of Laboratory, Fundacio Puigvert, Barcelona, Spain. $ Medical Staff, Unitat d’Estomatologia, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain. Q Chief of Unit, Unitat d’Hemostasia i Trombosi, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain. Address correspondence and reprint requests to Dr Fontcuberta: Unitat d’Hemostasia i Trombosi, Hospital de la Santa Creu i Sant Pau, Sant Antoni Maria Claret, 167, 08025 Barcelona, Spain 0 1996 American

Association

of Oral and Maxillofacial

Surgeons

0278-2391/96/5401-0006$3.00/O

27

28 Table

ORAL

1.

Patient

Characteristics Initial

No. of patients Mean age (mean t SD) Males Females Mitral or double prosthesis Aortic prosthesis Valvular heart disease

Study

64 59.7 -c 9.8 30 34 25 22 17

Group

5

28 56.3 + 9.4 12 16 14 7 7

the coagulation activity was unnecessary’ Moreover, reduction or discontinuation of the anticoagulant carries a potential thromboembolic risk in these patients.8 Local antifibrinolytic treatment has been used to reduce gingival bleeding after oral surgery in hemophilic patients.’ Recently, Sindet-Pedersen et al performed a study comparing the efficacy of tranexamic acid mouthwash with placebo.” We performed a prospective randomized study comparing several perioperative schedules using local antifibrinolytic agents to evaluate the need to reduce the level of anticoagulation before surgery. Patients

and Methods

Patients who were being treated with an oral anticoagulant agent because of heart disease (valvular heart disease or cardiac valve prosthesis), and who were to undergo oral surgery (tooth extraction), were enrolled in the study. They were recruited from February 1994 to October 1994. Before entering the study, informed consent was obtained from all. Exclusion criteria were previous thromboembolism under anticoagulant treatment; previous severe bleeding after tooth extraction before anticoagulant treatment; patients who preferred to go to their own surgeon; less than 3 months from the beginning of the anticoagulant treatment until oral surgery. Three patients did not want to proceed with the study once randomized, resulting in a cohort of 64 patients, 30 men and 34 women. Before their inclusion, and within a week before surgery, the effect of anticoagulant therapy was evaluated using the prothrombin time expressed in the International Normalized Ratio (INR)’ ’ so that all patients went to surgery with an INR between 2 and 3 if they had native valvular heart disease or with an INR between 2.5 and 4 if they had mechanical valves. To establish the level of anticoagulation at the time of surgery, another prothrombin time was performed immediately before the procedure. Random numbers were used to assign consecutive patients to treatment groups according to gender and to different causes for anticoagulation: mechanical prosthesis (47 patients) and valvular heart disease (17 patients) (Table 1).

SURGERY

IN ANTICOAGULATED

PATIENTS

When this study was concluded, the target levels of therapeutic anticoagulation were modified in our hospital. Recent publications suggest that less intensive levels are equally as effective in the prevention of thromboembolism in patients with mechanical prostheses.5 Therapeutic ranges are maintained between 2 and 3 for aortic prosthesis and cardiac valve disease and slightly reduced to 2.5 to 3.5 for a mitral prosthesis or double valvular substitution. Using these new criteria, 28 consecutive patients were added to the study without randomization and are referred to as group 5 (Tables 1, 2). All patients were instructed to communicate any bleeding complications after surgery. Surgery was performed by three dentists from the Unit of Stomatology of Sant Pau Hospital, and all of the procedures were supervised by the head of this unit. All patients underwent tooth extraction. Eightytwo were single-tooth extractions, and the other 10 were extraction of two adjacent teeth. Anesthesia was obtained by infiltrating 1.5 mL of a 3% solution of mepivacaine. In all patients, acenocoumarol (Sintrom; Geigy, Barcelona, Spain) was used for oral anticoagulant treatment. Patients were assigned to one of six groups of perioperative treatment as shown in Table 2. When acenocoumarol was reduced (groups 0, 1, and 2), calcium heparin (Calciparina; Sanofi, Barcelona, Spain) was administered. Epsilon-aminocaproic acid (EACA) (Caproamin Fides; Rottapharm, Valencia, Spain; 10 mL, 4 g per ampule) and tranexamic acid (TXA) (Amchafibrin Fides; Rottapharm, Valencia, Spain; 5 mL, 0.5 g per ampule) were the antifibrinolytics used. Prophylaxis against endocarditis with amoxicillin (Clamoxyl; Smith-Kline Beecham, Madrid, Spain) was begun preoperatively in all patients except those with an allergic history, in whom erythromycin (Pantomicina; Abbott Laboratories, Madrid, Spain) was used. Mouthwashes were performed with 1 ampule of antifibrinolytic agent for 2 minutes every 6 hours for 2 days. The parameters evaluated by the surgeons were gingival hypertrophy, local trauma performed during the procedure, and bleeding complications, which were evaluated according to the following classification: Mild bleeding: bleeding that stopped spontaneously or with minimal local compression Severe bleeding: bleeding that did not stop with the previous measures and required local adrenaline or continuous local compression using a gauze pack soaked in the corresponding antifibrinolytic agent until the bleeding stopped; major bleeding was also treated by reduction or supression of the oral anticoagulant.

Differences in treatment groups with respect to gender, types of heart disease, gingival hypertrophy, de-

SOUTO

29

ET AL

Table 2.

Schedules

Schedule GO

Gl

G2

G3 G4 G5

Used in the Study Acenocoumarol

Day Day Day Day Day Day Day Day Day Day Day Day

-2 -1 0 -2 -1 0 -2 -1 0 0 0 0

Half dose Half dose Usual dose Half dose Half dose Usual dose Half dose Half dose Usual dose Not modified Not modified Not modified

Heparin No 7,500 7,500 No 7,500 7,500 No 7,500 7,500 No No No

Antifibrinolytic

IW12 W/12

hrs hrs

III/12 IU/12

hrs hrs

III/12 IU/12

hrs hrs

Oral intake

of EACA

Tranexamic

acid as mouthwash

EACA

level of prosthesis < 2.5 for level of prosthesis > 3.5 for

anticoagulation: INR < 2 or valvular heart disease, mitral or double prosthesis. anticoagulation: INR > 3 or valvular heart disease, mitral or double prosthesis.

(4 g) before

surgery

as mouthwash

Tranexamic acid as mouthwash EACA as mouthwash Tranexamic acid as mouthwash

gree of surgical trauma, and number of teeth extracted were analyzed by the chi-squared test. Differences with respect to age and preoperative INR were analyzed with ANOVA and multiple contrasts by Duncan’s method.12 Differences between groups with respect to percentages of severe bleeding were analyzed by comparison of proportions. A logistic regression” was used to calculate the odds ratio (OR) associated with each group and the anticoagulation level. The latter was defined as: 1. Low aortic INR 2. High aortic INR

Immediately Cold water

irrigation

Cold water

irrigation

Cold water

irrigation

Cold water irrigation Cold water irrigation Tranexamic acid irrigation

not reduced preoperatively (G3 and G4) showed statistically higher INR levels than the other groups (GO, G 1, and G2) in which the dose of acenocoumarol was lowered preoperatively (CI of the difference, 0.15 to 1.21). In group 5, in which the therapeutic range of anticoagulation was slightly lower, statistically significant differences were found with groups G3 and G4 (CI of the difference, 0.15 to 1.06), but not with groups GO, Gl, and G2. Figure 2 reflects the percentages of hemorrhagic

for and for and

INR 61

.

. .

Patients in group 0 with good anticoagulation were considered as the baseline situation, and all of the odds ratios were calculated with this premise. All relevant data were expressed in the 95% confidence interval (CI), and a value <.05 was considered significant. A standard statistical program (SPSS-PC, 5.1) was used for all the calculations and was performed on a 486SX computer.

.

P

@ . *

.

. .

.

.

:.

.> ‘**

. :

A

Results

2

l* .

l *.* .

l.

Some of the clinical characteristics of the patients are shown in Table 1. There were no significant differences between groups with respect to gender, age, cause for anticoagulation, gingival hyperthrophy, surgical trauma, or number of extracted teeth. None of these variables increased the hemorrhagic risk. Figure 1 shows the number of patients included in each treatment group as well as the INR levels preceding the surgical procedure. The groups in which the anticoagulant treatment was

Postoperative

. *** .

3

.

. .

.

4 -e..a* ..

.

11 GO

Gl

G2

G3

G4

G5

n

13

12

14

12

13

28

mean

2.50

2.93

2.50

3.29

3.50

2.82

FIGURE procedure.

1.

INR

levels

of patients

immediately

before

surgical

30

ORAL

Table 3. Treatment

Schedule GI G2 G3 G4 G5

FIGURE 2. ment group.

Percentage

of hemorrhagic

complications

SURGERY

IN ANTICOAGULATED

Hemorrhagic Regimens

Risk With

Crude Odds Ratio

INR Adjusted Odds Ratio

5.5 (0.8-36.2) 3.05 (0.6-22.8) I.10 (0.13-9.34) 2.44 (0.36-16.5) 0.20 (0.02-2.48)

4.95 (0.63-39.3) 3.17 (0.45-22.5) 0.88 (0.09-8.75) I .64 (0.20-13.5) 0.12 (0.01 -I .94)

PATIENTS

Different

High Anticoagulation 15.8

LOW Anticoagulation I .I2

6.83

2.26

1.26

0.38

2.64

0.44

0.16

0.12

in each treat-

complications. Globally, there was a statistically significant difference between the different groups (chisquared = 13.79, P = .017). No significant differences were observed between Gl and G2, between G3 and G4, or between GO and Gl + G2 + G3 + G4 when considered together. However, significant differences were observed between G5 and Gl + G2 (CI of the difference, 18.5 to 58.9%) and between G5 and G3 f G4 (CI of the difference, 2.3 to 38.5%), but not between GO and G5. The level of anticoagulation during the procedure influenced the hemorrhagic risk; in poorly anticoagulated patients the odds ratio was 0.15 (CI, 0.03 to 0.7 l), whereas in highly anticoagulated patients it was 3.94 (CI, 1.17 to 13.3). Moreover, 7.1% of poorly anticoagulated patients (CI, 0.9 to 23.5%) had severe bleeding, in contrast with 47.1% (CI, 23 to 72%) of highly anticoagulated patients with this complication. Thus, the anticoagulation level was a confusing variable capable of modifying the effect of the schedule on bleeding. For this reason, we adjusted the calculation of the OR accordingly. The remaining analyzed variables (age, gender, and reason for anticoagulation) did not show any relation with hemorrhagic risk, although age and gender were included in the regression model to calculate adjusted OR. Table 3 reflects the OR (crude and adjusted to INR) of the different schedules. Gl and G2 had the highest OR of severe bleeding, G3 and G4 were quite similar to GO (reference group), and G5 showed a protective effect. When we distinguished between poorly or highly anticoagulated patients, these results were more evident, and there was a protective effect in the schedules that did not modify acenocoumarol dose or use heparin (G3, G4, and G5). At the same time, schedules Gl + G2 when analyzed together showed a higher risk

of severe bleeding compared with G3 + G4, GO, and G5 (Table 4).

Discussion In the past, recommendations for the administration of oral anticoagulants in conjunction with oral surgery have varied, ranging from complete withdrawal of the anticoagulant’4 to the maintenance of an unchanged therapeutic leve1.‘5,‘6The discontinuation of anticoagulant therapy involves a risk of thromboembolism, especially in patients with prosthetic heart valves.’ The effect of the antifibrinolytic treatment in oral surgery may be explained by the presenceof activators of fibrinolysis and the absenceof physiologic inhibitors of fibrinolysis in saliva.17 The concentration of TXA

Table 4. Association

Comparison of Similar

of Hemorrhagic Schedules

Al1 Patients Crude OR

GI fG2 vs G3 + G4 GI +G2vs GO Gl

+G2vs G5

G3 + G4 vs GO G3 + G4 YS G5

Adjusted OR

Selected High Anticoagulation

Risk After

Patients LOW Anticoagulation

2.32 (0.7-7.7)

5.89 (1.4-31.9)

6.45

8.62

4.03 (0.7-22)

4.14 (0.7-26)

10.15

2.04

9.29 (2.3-30.8)

10.37 (2.8-61.8)

24.5

8.7

I .74 (0.3-10.1)

1.15 (0.2-8.0)

I .96

0.38

4.00 (0917.7)

2.59 (0.6-13.8)

4.09

I .42

GO was the reference schedule. Gl and G2 modified the dose of acenocoumarol, and G3 and G4 did not. G5 group was not randomized, and the dose of acenocoumarol was not modified.

SOUTO

31

ET AL

after oral administration (5 mL solution) is not sufficient to suppress local fibrinolysis. However, the concentration of TXA in saliva after mouth rinsing four times a day was sufficient to reduce the amount of lysed fibrin’* so that the incidence of postoperative bleeding complications was significantly reduced. Other studies have demonstrated that maintenance of oral anticoagulant therapy in conjunction with oral surgery does not result in severe bleeding complications in patients receiving tranexamic acid when compared with placebo. ” In the study by Sindet-Pedersen et al, oral surgery was performed with no change in the level of anticoagulant therapy, and treatment with the antifibrinolytic agent, TXA, was continued for 7 days after surgery.‘O Our study initially compared five treatment schedules. Group 0 followed our previously used routine schedule, which consisted of reducing oral anticoagulation before surgery and adding subcutaneous heparin and oral EACA. Gl and G2 were similar to GO with respect to anticoagulation therapy. In these groups, we wanted to evaluate the local effect of EACA and TXA in mouthwashes. Acenocoumarol was not modified in G3 and G4, and EACA and TXA were compared under the same conditions. It is difficult to analyze differences between TXA and EACA because Gl had a higher risk than G2 but, on the contrary, G4 seemed better than G3. Moreover, the absence of significant differences in the OR associated with these four schedules makes it impossible to judge which antifibrinolytic is preferable. We did not observe significant differences in the percentage of severe bleeding or hemorrhagic risk (expressed in odds ratio) with respect to GO (Table 3) in Gl, G2, G3, and G4. However, a higher odds ratio was observed for Gl and G2, schedules in which the oral anticoagulation was reduced. Although we do not find an explanation for these results, it is probable that the patients included in Gl and G2 were more anticoagulated by the combination of acenocoumarol and heparin, an effect not indicated by the INR. If highly anticoagulated patients are selected, we can see differences between schedules more clearly. High hemorrhagic risk is associated with Gl and G2. Logically, patients with an INR above the desired therapeutic range and also under heparin treatment have a higher risk of severe bleeding complications. However, it is surprising that schedules Gl and G2, although very similar to the reference group, showed a higher risk. The explanation could be related to some unknown effect of EACA administered orally in GO. Group 5 was evaluated under different conditions once the prior study was finished. The main difference was the INR level before surgery, because we lowered the intensity of anticoagulation in our patients. For this

reason, this schedule is not comparable with the rest. However, we decided to include it to evaluate its protective effect (Tables 3,4). This effect was independent of the anticoagulation level; ie, with this schedule of treatment, the excessive anticoagulation did not increase the hemorrhagic risk. In this schedule, TXA was used as an antifibrinolytic agent but, unlike in G3, was immediately applied locally on the surgical wound. Unlike the study by Sindet-Pedersen et al,‘O in which mouthwashes with TXA were performed for 7 days after surgery, in our schedule 5, a low incidence of bleeding was observed with only 2 days of postoperative mouthwashes. We are aware of the limitations of this study, especially related to the small size of the treatment groups and to the lack of comparability of scheduleG5. Nevertheless,we think that reduction of anticoagulation levels, according to the latest recommendations,5 could decreasethe hemorrhagic risk. Furthermore, the use of TXA decreasesthis risk independent of the INR. Since this study was finished, oral surgery has been done in more than 100 patients chronically anticoagulated for several reasons.All have been treated following schedule5, and none has had major bleeding complications. Our results with the schedulesthat maintain the oral anticoagulant regimen and use local antifibrinolytic agents are comparable to those that diminish acenocoumarol treatment with respect to severe bleeding complications, and they offer the advantage of decreasing the thromboembolic risk. If TXA is applied on the wound immediately after surgery and followed by mouthwashes for 2 days, the hemorrhagic risk is low, and therefore this model seems to be the most desirable. We believe that heparin administration is a noncontrollable risk factor and carries a higher degree of complexity and discomfort for the patient. References 1. Duveau D: Anticoagulation is necessary in all patients with mechanical prosthesis in sinus rhythm. Z Kardiol 75:326, 1986(suppl 2) 2. Stein PD, Collins JJ, Kantrowitz A: Antithrombotic therapy in mechanical and biological prosthetic heart valves and saphenous vein bypass grafts. Chest 89:46S, 1986(suppl 2) 3. Levine HJ, Pauker SG, Salzman EW: Antithrombotic therapy in valvular heart disease. Chest 89:368, 1986(suppl 2) 4. Mok CK, Boey J, Wang R, et al: Warfarin versus dipyridamolaspirin and pentoxifylline-aspirine for the prevention of prosthetic heart valve thromboembolism: A prospective clinical trial. Circulation 72: 1059, 1985 5. Hirsch J: Oral anticoagulant drugs. N Engl J Med 324:1865, 1991 6. Kelton JG, Hirsch J: Bleeding associated with antithrombotic therapy. Semin Hematol 17:259, 1980 7. DeClerck D, Vinckier F, Vermylen J: Influence of anticoagulation on blood loss following dental extraction. J Dent Res 711387, 1992 8. Ogiuchi H, Ando T, Tanaka M: Clinical reports on dental extrac-

32

9. 10. 11.

12. 13.

DISCUSSION tion from patients undergoing oral anticoagulant therapy. Bull Tokyo Dent Co11 26:205, 1985 Sindet-Pedersen S, Stenbjerg S, Ingerslev J: Control of gingival hemorrhage in hemophilic patients by inhibition of fibrinolysis with tranexamic acid. J Periodont Res 23:72, 1988 Sindet-Pedersen S, Ramstrcim G, Bemvil S, et al: Hemostatic effect of tranexamic acid mouthwash in anticoagulant-t patients undergoing oral surgery. N Engl J Med 320:840, 1989 Loeliger EA: International Committee for Standardization in Haematology. International Committee on Thrombosis and Haemostasis. ICSHKTH recommendations for reporting prothrombin time in oral anticoagulant control. Thromb Haemost 53:155, 1985 Riba MD: Modelo Lineal de ANOVA. Barcelona, Spain, Herder, 1990 Kleinbaum DG: Modeling: Theoretical considerations, in van

J Oral Maxillofac 54:32-33, 1996

14. 15. 16. 17. 18.

Nostrand Reinhold (ed): Epidemiologic Research: Principles and Quantitative Methods. New York, NY, 1982, p 419 Ziffer AM, Scopp IW, Beck J, et al: Profound bleeding after dental extractions during dicumarol therapy. N Engl J Med 256:351, 1957 Anavi Y, Sharon A, Gutman D, et al: Dental extractions during anticoagulant therapy. Isr J Dent Med 28:9, 1981 Bailey BM, Fordyce AM: Complications of dental extractions in patients receiving warfarin anticoagulant therapy. Br Dent J 155:308, 1983 Sindet-Petersen S, Gram J, Jespersen J: Characterization of plasminogen activators in unstimulated and stimulated human whole saliva. J Dent Res 66:1199, 1987 Sindet-Pedersen S: Distribution of tranexamic acid to plasma and saliva after oral administration and mouth rinsing: A pharmacokinetic study. J Clin Pharmacol 27:1005, 1987

Surg

Discussion Oral Surgery in Anticoagulated Patients Without Reducing the Dose of Oral Anticoagulant: A Prospective Randomized Study Steen Sindet-Pedersen, Michael Vaeth, PhD

DOS, Dr Med Sci

Aarhus University and University Hospital, Denmark

This article hasseveralpoints worthy of discussion.Regardingthe designof the study, we find it confusingthat the title and the abstract describethe study as a prospective randomized study although only 70% of the participants were included in the randomization.There is alsono information about blinding of the study medicationsor patient compliancewith respectto taking the study medications. The study includessomevery smallgroupsof patientsand we believe that this reducesthe possibilitiesof drawing useful conclusionsfrom the data. In addition, information on the distribution of diagnosesin the differently medicated groupsis missing.Also, the effect on the coagulationsystem of administeringstandardizeddosesof heparinto group 0, 1, and 2 hasnot beenmonitoredby appropriatelaboratory analyses.The authorsmentionthat they evaluatedsome parameters,including gingival hypertrophy and local trauma which were performedduring the procedure,etc. However, they do not use these parametersin their analysisof the outcomeof the study. Regardingthe antifibrinolytic treatment, the information about the numberof dosesand duration of treatmentis unclear.We are alsoworried about the fact that thrombosis-prone patients(ie, group 0) in the present study were treatedwith a systemicallyadministeredantifibrinolytic drug (EACA). This is not clinical practicein our part of the world basedon the conceptthat a systemicreduction of fibrinolysis may increasethe risk of thrombosis;this conceptconstitutesthe whole basisfor implementinglocal antifibrinolytic treatment among patients under treatment with oral anticoagulants.

In respectto the statisticalanalysis,accordingto Table 2, groupsGl, G2, G3,andG4 constitutea simplefactorialexperimentin which the treatmentallocationcan be organizedin a two by two table.A separateanalysisof thesegroupswould facilitatea simpleinterpretationof the data.By includinggroup 0 and group5 in the analysis,the interpretationof the results becomesmuchmorecomplex,and this problemshouldhave heenaddressed in thedesignphaseof the study.Thedescription of the statisticalmethodsusedis alsotoo brief to presentfull understanding of theresultsthat arepresented later.Concerning thelogisticregression analysis,we learnthat “patientsin group 0 with goodanticoagulationwere consideredas the baseline situation.” Doesthis meanthat all patientsin group 0 were includedin the referencegroupor were patientswith “poor” anticoagulation excluded? The authorsperforma globalcomparison of all six treatment groups.It would, however, seemrelevant alsoto considera globalcomparisonof the five treatmentsthat wereincludedin therandomizedpartof theinvestigation.The authorsthencompare the individual groupsand combinationof groups.Of course,this procedureassures that the comparisons arebased on a largenumberof patients,but it impairsthe interpretation of the results,and it is also lessclear which questionsthe authorswishto answerwith theseanalyses.Moreover, we are missinginformationabout complicationsthat developed,the severity of thesecomplications,how they were treated,and whethertherewere any severeirreversiblecomplications. In the descriptionof theimportanceof thelevel of anticoagulationto the rate of complications, it is unclear how the statistical analysiswasdone.For instance,the effect of the level of anticoagulation on the complicationrate is described,but it is unclearhow the oddsratio for poorly anticoagulated patients andthe oddsratio for highly anticoagulated patientsarecalculatedagainstgroup0, or partsof group0, asthereferencegroup. It is not mentionedwhetherthe regression analysisallowsfor the fact that thereare five different treatmentgroupsbesides the referencegroup. It would be moreinterestingto describe the effect of the level of anticoagulationby calculatingan oddsratio for highly anticoagulated patientsrelative to poorly