Documente Academic
Documente Profesional
Documente Cultură
DOI 10.1007/s00784-013-1133-x
ORIGINAL ARTICLE
Received: 27 April 2013 / Accepted: 31 October 2013 / Published online: 28 November 2013
# Springer-Verlag Berlin Heidelberg 2013
obturation of spongious bone with bony wax (B. Braun postoperative bleeding between 1998 and 2009. Most of the
Melsungen AG, Melsungen, Germany). Other hemostatic patients had been treated in dental offices during the day and
methods require additional equipment. Thermal modalities presented with delayed bleeding during nighttime. Since these
include coagulation with diathermy or laser. Biochemical patients registered in our clinic due to secondary bleeding, a
means include human fibrinogen, human thrombin, and the common clinical standard had not been applied. Some of these
application of hemostatic agents such as Tabotamp patients might have had primary wound closure but most of
(Ethicon Biosurgery, Johnson & Johnson MEDICAL themespecially due to the fact that the majority had been
GmbH, Norderstedt, Germany), an oxidized regenerated cel- extractions onlyhad none. A total of 4,946 patients were
lulose that also exerts bactericidal effects [4]; Spongostan, a reviewed for sex, age, location of bleeding, the presence of
gelatine sponge of porcine origin; fibrin glue; TachoSil anticoagulation or congenital blood disorders, the underlying
(Nycomed International Management GmbH, Zurich, Swit- surgical intervention, inpatient or outpatient, the date of sur-
zerland), an equine-derived, honeycomb-like collagen coated gery, and the date when bleeding was registered. Drugs other
with a dry layer of coagulation factors [5]; and tranexamic than anticoagulants were not evaluated. Only patients who
acid (Cyklokapron, MEDA Pharma GmbH, Wangen- originally underwent planned oral surgical interventions were
Brttisellen, Switzerland), a synthetic derivative of the amino included. Patients with oral pathologies as sources of bleeding,
acid lysine and an antifibrinolytic drug [6]. such as periodontitis, trauma, or tumors were excluded, as well
To stop secondary bleeding, the same intraoperative treat- as patients with acute interventions (abscess incision). Cases
ment modalities can be utilized. with missing or flawed documentation were also excluded.
The mechanisms of bleeding have to be distinguished This study protocol was sent to and approved by the Ethical
between primary and secondary bleeding. Primary bleeding Committee of the Medical University Vienna (EK# 371/2010).
occurs due to direct trauma caused by a surgical instrument or Of the 4,946 cases with postoperative bleeding complica-
the rupture of tissue, such as when periodontal fibers are torn tions, 3,113 were excluded: 1,098 with abscess incision, 566
apart during tooth extraction or during blunt scissor prepara- with dentoalveolar bleeding due to trauma, 112 with bleeding
tion. A secondary bleeding can result from the ease of the due to periodontitis, 88 with bleeding from tumor erosion, 308
initial vascular spasm due to epinephrine added to the local with an unspecified surgical procedure, 697 due to missing or
anesthesia, or by injury to the soft and fragile blood clot. invalid date of surgery, and 244 without documented surgical
Numerous other reasons of secondary bleeding are also pos- treatments. Data of the 1,833 eligible patients were
sible, including external heat, rising blood pressure, adaption reformatted into a Microsoft Excel spreadsheet.
of coagulation therapy, and hyperemia of the inflamed wound.
This retrospective study examines a 12-year cohort of
patients, between 1998 and 2009, who suffered from postop- Data analysis
erative bleeding requiring hemostatic intervention. The aims
of the study were (a) to point out the reasons for postoperative Data was analyzed with statistical methods using the open source
bleeding, (b) to examine the time interval between surgery and statistical software package R (Version 2.15.1, The R Foun-
onset of bleeding, and (c) to give recommendations for the dation for Statistical Computing, http://www.r-project.org).
best time for operative interventions and to identify patients at Possible factors (independent parameters) influencing the
risk of further inpatient treatment. time interval until the onset of bleeding (dependent variable)
were analyzed. Since the bleeding interval is not normally
distributed but rather count data (starts with zero, never neg-
Materials and methods ative, and always an integer), a Poisson distribution or even
better a negative binomial (NegBin) distribution may be ap-
The University Hospital for Cranio-, Maxillofacial and Oral plied. The interval was estimated using negative binomial
Surgery sees a substantial proportion of patients with delayed count regression models (NegBin II) since the variance of
hemorrhage after oral surgery. This is to the greatest possible bleeding time was greater than the mean (i.e., overdispersed).
extent because the University Hospital for Cranio-, Maxillofa- Univariate NegBin II models were compared to the null by
cial and Oral Surgery is the only faculty available in eastern means of likelihood ratio testing. Finally, the full multivariate
Austria to treat patients after 1 a.m. Consequently, the University NegBin II model was computed and compared to individual
Hospital for Cranio-, Maxillofacial and Oral Surgery has access models from likelihood ratio tests. The significance of each
to a rich database of patients for a retrospective study of post- regressors estimate was calculated using Walds test and the
operative bleeding. The present study includes patients since the total effect using likelihood ratio testing. In case of nominal
implementation of the clinics digital documentation system in factors (such as localization), dummy variables are introduced
1998 (Clinicware, Version 336.026, GWI AG, Trier, Germany). whereby four regions (incisor, canine, premolar, molar) are
The study included patients who presented with delayed modelled by three dummy variables: canine, premolar, and
Clin Oral Invest (2014) 18:16551661 1657
molar. The model then reports the bleeding interval of incisors on anticoagulant. No anticoagulation was used as the refer-
as constant value and the differences of canine/premolar/mo- ence group. The comparison of anticoagulation and congenital
lar region as estimate of the dummy variable. blood disorders to no anticoagulation showed a significant
difference in the time between intervention and onset of the
bleeding (likelihood ratio test for maximum likelihood estima-
Results tion (MLE) method: p LRT =0.00017), whereby phenprocoumon
and congenital disorders showed a similar significant effect of
Out of the 4,946 patients with postoperative bleeding, 1,833 1.80 (e0.589, p Walds test <104) and 1.84 (e0.608, p Walds test =0.024)
patients were eligible according to the inclusion and exclusion longer mean time, respectively. The effect of acetylsalicylic acid
criteria. An overall bleeding interval (period between date of was smaller and shown to be insignificant (e0.167 =1.18,
surgery and date of presentation for postoperative bleeding) of p Walds test =0.338). Patients on clopidogrel started bleeding ear-
1.4 days and a variance of 12.1 was observed ranging from 0 lier than the reference group (e0.709 =0.492, p Walds test =0.199),
to 37 days. A total of 64.97 % of all patients showed delayed but the difference was also statistically insignificant. Similarly,
bleeding on the day of surgery, 20.96 % in the first 3 days there was an insignificant difference between patients on multi-
following surgery, and 9.18 % between the fourth and the ple anticoagulants and the reference group (e0.445 =1.56,
seventh day after surgery. p Walds test =0.550).
Anticoagulants Of the 1,833 cases, 1,101 (60.1 %) had no Surgery Of the 1,833 patients, 1,746 (95.3 %, bleeding inter-
history of anticoagulants. Of the remaining 732 patients, 394 val=1.33.5; range 037) suffered bleeding after tooth ex-
(21.5 %, bleeding interval=23.7) had taken phenprocoumon, tractions, 29 patients suffered bleeding (1.6 %, bleeding inter-
233 (12.7 %, bleeding interval=1.33) had taken acetylsalicylic val=0.41.0; range 05) after failed extractions, 34 patients
acid, 33 (1.8 %, bleeding interval= 0.8 1.7) had taken (1.9 %, bleeding interval=2.43.2; range 012) after dental
clopidogrel, 17 (0.9 %, bleeding interval=1.32.1) had more implantation, 13 patients (0.7 %, bleeding interval=2.95 ;
than one type of anticoagulant, and 78 (4.3 %, bleeding inter- range 015) following cystectomies or root-end resections,
val=23.3) had congenital blood disorders. Bleeding intervals and 11 patients (0.6 %, bleeding interval=11.2; range 03)
of patients under phenprocoumon and acetylsalicylic acid are after other dental treatments. The type of surgery did not
given in Fig. 1. All 1,833 patients contributed to the significantly influence the bleeding interval in the NegBin II
anticoagulation NegBin II model regressing bleeding interval model (likelihood ratio test for MLE method: p LRT =0.054).
interval comparing
phenprocoumon and
acetylsalicylic acid
8
Days after surgical intervention
4 2
0 6
aspirine phenprocoumon
Anticoagulation
1658 Clin Oral Invest (2014) 18:16551661
The postoperative bleeding interval from tooth extractions (bleeding interval 0.71.4, 05), 12.1 % affecting the maxil-
was used as the reference group. The postoperative bleeding lary premolar region (bleeding interval 3.27.6, 037), and
interval in cystectomies was longer than tooth extractions, but 14.7 % affecting the mandibular premolar region (bleeding
this difference was found to be insignificant (factor=e0.777 = interval 1.93.4, 014) (Fig. 2). The localization was shown
2.18, p Walds test =0.224). Bleeding interval after dental implan- to be insignificant in affecting bleeding interval (likelihood
tation was also insignificantly different from tooth extractions ratio test for MLE method: p LRT =0.093). However, when
(factor=e0.597 =1.82, p Walds test =0.135), as was dental proce- using maxillary incisors as reference group for comparison,
dures (preparation) when compared to tooth extractions maxillary canines (factor=e1.245 =3.47, p Walds test =0.046) and
(e0.295 =0.74, p Walds test =0.688). In contrast, failed extraction premolars (factor=e1.390 =4.01, p Walds test =0.021) showed sig-
showed a significantly smaller bleeding interval when com- nificantly longer bleeding intervals according to the univariate
pared to successful tooth extractions (factor=e1.178 =0.31, NegBin II model. There were no significant differences in
p Walds test =0.020). other groups (maxillary molars 2.01-fold, p Walds test =0.171;
mandibular incisors 1.08-fold, p Walds test =0.913; mandibular
Sex Of the 1,833 patients, 1,078 patients were male (58.8 %, canines 0.88-fold, p Walds test =0.866; mandibular premolars
age 48.620.7, ranging from 2.7 to 97.8 years) and 755 were 2.44-fold, p Walds test =0.135; mandibular molars 1.97-fold,
female (41.2 %, age 50.522.0, ranging from 0.2 to 98.0 years). p Walds test =0.187).
A mean bleeding interval of 1.33.4 (range 037) was ob- All factors which were significant in the univariate statis-
served in female patients and 1.43.5 (ranging from 0 to 26) in tics (anticoagulation and clinic) were also significant in the
male patients. The univariate NegBin II model analysis of multivariate statistics (see Table 1).
cohort sex did not show significant differences in bleeding
intervals between males and females (likelihood ratio test for
MLE method: p LRT =0.265; e0.127 =1.14, p Walds test =0.264).
Discussion
Age The bleeding interval insignificantly lengthened by a factor
of 1.004 for every yearly increase in age (=e0.004, p Walds test = In countries of overaging population, the rising number of
0.124, p LRT =0.128). patients treated with anticoagulants requires intensified col-
laboration between internists and dentists. Bleeding after an
Clinic Inpatients and outpatients showed a mean bleeding oral (surgical) intervention is a common and unsettling com-
interval of 0.10.8 (ranging from 0 to 8 days) and 1.43.6 plication, about which every patient has to be informed before
(ranging from 0 to 37), respectively. This difference between treatment [7, 8]. Various factorsespecially anticoagulant
inpatients and outpatients was shown to be significant accord- therapies such as phenprocoumon etc. and congenital blood
ing to the univariate NegBin II regression model (factor= disordersincrease the risk of bleeding and especially de-
e2.39 =0.09, p Walds test <1012, likelihood ratio test for MLE layed bleeding [9]. In this study, all patients with delayed
method: p LRT <1012). bleeding after projectable dental surgical interventions (i.e.,
with possibility of preoperative anticoagulant therapy adap-
Jaw Of the 1,833 patients, 763 had a valid specification of the tion) were included to provide further evidence of the time
jaw and 618 had localization of teeth. Bleedings were almost interval between surgery and onset of bleeding. Cases of
equally distributed to the maxilla (n =401, 52.6 %, bleeding spontaneous bleeding due to trauma, inflammation, or tumor
interval=1.94.6, range 037) and the mandible (n =362, were excluded (in total, 19.9 % of the cohort with hemostatic
47.4 %, bleeding interval=1.53.4, range 033). In the measures). As a possible source of bias, 65.5 % of the patients
NegBin II regression model (likelihood ratio test for MLE with previous dental surgeries were included while the re-
method: p LRT =0.159), bleeding in the maxilla started slightly maining 34.5 % had invalid documentation.
later than the mandible but the difference was found to be
insignificant (factor=e0.232 =1.26, p =0.158). Bleeding interval Independent of any factor, bleeding oc-
curred on the same day of surgery in more than two thirds of
Localization In both maxilla and mandible localization, bleed- the cases. It can be assumed that the majority of postoperative
ing was mainly registered in the molar region (maxilla 70.6 %, recurrent bleedings are persisting bleedings temporary sup-
bleeding interval 1.63.7, 031; mandible 69.3 %, bleeding pressed by adrenaline added to local anesthetics since most
interval 1.64, 033). Only 7.4 % of bleedings occurred in the bleeding complications occur within few hours after surgery.
maxillary incisor region (bleeding interval 0.81.4, 04),
9.2 % in the mandibular incisor region (bleeding interval 0.9 Cause It is an underestimated fact that the main surgical
1.4, 06), 9.9 % affecting the maxillary canine (bleeding reason for a bleeding complication was a simple tooth extrac-
interval 2.86.8, 036), 6.8 % affecting the mandibular canine tion (95.3 %)not advanced dentoalveolar surgery. Typically,
Clin Oral Invest (2014) 18:16551661 1659
10
interval comparing incisors,
canines, premolars, and molars
8
Days after surgical intervention
6
4
2
0
I C PM M
Localization
a conventional tooth extraction does not require soft tissue compression in the molar region is more challenging for the
wound closure, hence correct placement of the cotton swab patient himself. Exposed bone surface has to be considered in
and adequate time of compression is crucial. Posterior teeth planning of the operative procedure.
with more than one andmoreovercontorted roots leave
large bony wounds, leading to more than two thirds of bleed- Therapy with anticoagulants In this context, the anticoagulant
ing complications in the molar region. Additionally, therapy plays an important role. Although two thirds of the
Factor Subfactor Estimate Std. error eestimate z value p Walds test p LRT
Estimates of the coefficients, standard errors, exponential of coefficient, and z value are given from left to right. Probability values of Walds test and
likelihood ratio test are given in the rightmost columns. Null deviance, 575.79 on 615 degrees of freedom; residual deviance, 445.64 on 600 degrees of
freedom; AIC, 1,795.1; theta, 0.2948; std. error, 0.0281; 2log-likelihood, 1,761.1060
1660 Clin Oral Invest (2014) 18:16551661
patients had no history of anticoagulants, the second largest Table 2 Interaction of chemotherapeutic agents and potency of
phenprocoumon
group was taking phenprocoumon with significant longer
surgery-to-bleeding intervals. Since the study period ended in Elevating the anticoagulant effect: Inhibiting the anticoagulant
2009, the new anticoagulants such as rivaroxaban (Xarelto) effect:
and dabigatran (Pradaxa) had not been considered in our
Allopurinol 6-Mercaptopurine
study. Experience with the bleeding risk of this drug group
Aminoglycoside Azathioprine
and recommendations on how to handle those before or during
Amiodarone Barbiturates
oral surgical interventions are under current investigation.
Anabolic steroids Carbamazepine
Hemostasis is a finely tuned orchestration of physical and
Cephalosporins Chlordiazepoxide
biochemical forces to arrest bleeding. The trauma induced by
Chinin Cholestyramine
tooth extraction leads to blood vessel injury and extravasation
of blood to the surrounding tissue. The endothelium lining of Chinidin Corticosteroids
vessels is physically denuded by injury, thereby gets activated Chinolone Diuretics
and generates a highly pro-thrombotic interface. Circulating Cimetidine Glutethimide
platelets bind and get activated on the injured site as a conse- Cisapride Griseofulvin
quence, while the coagulation pathway is triggered by the Clofibrate and other fibrates Nafcillin
exposure to tissue factor (TF). Subendothelial TF binds factor Chloramphenicol Phenytoin
VIIa, and the newly formed complex can convert factor X to Cloxacillin Rifampicin
Xa, which in turn assembles with cofactor Va and Ca2+ on the Co-trimoxazole Sucralfate
platelet phospholipid membrane to form a prothrombinase COX-2 inhibitors Trazodone
complex that cleaves prothrombin into active thrombin. Disulfiram Vitamin K
Thrombin production is sustained by an alternate tenase, the Erythromycin
complex of VIIIa and IXa, and further amplified through Fluconazole
positive feedback by thrombins activation of VIIIa and Va, Glucagon
as well as XIa required for factor IX activation. Finally, Isoniazid
thrombin cleaves fibrinogen and activates factor XIII to en- Macrolide
ables fibrin cross-linking. Cross-linked fibrin binds platelet Metronidazole
clusters, leading to hemostatic plug formation in the physical Miconazole
barrier to limit the blood loss after tooth extraction. Throm- Omeprazole
bins ability to activate inflammatory mediators and platelets Orlistat
capability to secrete pro-inflammatory and immune mediators Phenylbutazone
such as serotonin, platelet factor 4 (PF4), P-selectin, and Piroxicam
CD40L (CD154), as well as their ability to express toll like Propafenone
receptors (TLR) 4 and 9, suggest a bridge between their Propranolol
function in hemostasis and immunity. The results suggest that Salicylates
while the inhibition of platelets has a short-term effect on Statins
wound healing, the inhibition of clotting factors delays wound Sulfinpyrazone
healing with prolonged risk of postoperative bleeding. Sulfonamide
Since simple withdrawal of the anticoagulant therapy before Tamoxifen
oral surgery might induce a rebound effect and cause throm- Tetracycline
boembolic events in patients suffering from atherosclerosis Triazol derivatives
[10, 11], it is recommended that anticoagulant therapy is ad- Virustatics (saquinavir)
justed [12] but not stopped [11]. Several preoperative proto- Selective serotonin reuptake inhibitors
cols have been established to prevent postoperative bleeding: Immunosuppressant (leflunomide)
(1) adjusting the international normalized ratio (INR) to upper
limits of 3.0 [13, 14] when phenprocoumon is administered,
(2) changing to low molecular weight heparins, (3) pretreating It can be assumed that preoperative and operative manage-
with antibiotics in the case of inflammation, or (4) substituting ment is planned more carefully in patients with more than one
coagulation factors, when blood disorders are known. Previous anticoagulant drug as there are not even 0.9 % of registered
recommendations to discontinue phenprocoumon for smaller bleeding events. In addition, medication interactions have to be
surgical interventions have become obsolete in the light of the considered, leading to numbers of up to one fifth of bleeding
elevated risk of severe thromboembolic events, especially complications in patients treated with phenprocoumon.
when the target INR is 2.0 to 3.0 [15, 16]. Anticoagulation effects and thus bleeding susceptibility of
Clin Oral Invest (2014) 18:16551661 1661