Bleeding complications in skin cancer surgery are associated with warfarin but not aspirin therapy - a prospective study.
Anthony J. Dixon, MBBS FACRRM FACSCM
Mary P. Dixon, B Appl Sci (Nursing)
John B. Dixon, MBBS FRACGP PhD
Method: This was a prospective study of 5950 skin lesions excised on 2394 patients. No patient stopped taking aspirin or warfarin unless the international normalized ratio (INR) exceeded 3.0.
Objective: The aim was to identify risk factors for post operative bleeding following skin cancer surgery.
Results: The rate of post operative bleeding was 0.7 per cent overall and 2.5 per cent in the 320 patients taking warfarin. The rate of bleeding was 1.0 per cent for skin flap repairs, 0.4 per cent for simple excision and closure, and 5.0 per cent for skin grafts.
Diabetic patients and smokers were not at increased risk of bleeding. There were four independent risk factors for bleeding: age 67 years or older, odds ratio (OR) 4.7 (95per cent confidence interval, 1.8 to 12.2); P=0.002, warfarin therapy (OR 2.9, (1.4 to 6.3); P=0.006), surgery on or around the ear (OR 2.6 (1.2 to 5.7); P=0.012) and closure with a skin flap or graft (OR 2.7 (1.4 – 5.3); P=0.004). Aspirin therapy was not an independent risk factor for bleeding.
Conclusion: Warfarin therapy, increasing age, surgery on or around the ear and closure with flap or graft are independent risk factors for post operative bleeding complications in skin cancer surgery. Most postoperative bleeds were inconvenient but not life threatening, unlike the potential risk of thromboembolism after stopping warfarin or aspirin. There was no case for discontinuing aspirin before skin surgery, but the INR should be monitored in patients taking warfarin.
Some patients who require minor surgery for skin cancer will be taking antithrombotic medication such as aspirin or warfarin. Their management varies greatly from surgeon to surgeon; some routinely stop these drugs before surgery, whereas others continue them.105
Several studies have reported no increased risk of bleeding associated with continuing warfarin or aspirin. All but one of these studies were very small, with fewer than 100 patients on aspirin or non steroidal anti-inflammatory (NSAID) and up to 16 patients on warfarin67, 106, 109, 198. The largest involved 286 patients taking aspirin or NSAIDs and 26 taking warfarin107. The power of these studies would have only allowed the detection of a large difference in bleeding risk and would have been limited by other confounding risk factors. Bleeding is the commonest complication of skin surgery (incidence of 3 per cent), and use of anticoagulants or immunosuppressant drugs is a risk factor.44
In recent years surgeons have been encouraged to continue warfarin and / or aspirin therapy before excision of skin tumours106, 107, 109, 155, 196, 199, 200, 204-206, accepting an international normalized ratio (INR) of up to 3.5.204 Stopping antithrombotic drugs risks life threatening thromboembolic sequelae, even when medication is stopped only for a short period. 105, 193, 201, 207.
Other forms of surgery have some data available to guide surgeons in their decision whether or not to stop antithrombotic medication. There is mixed advice regarding continuing warfarin and aspirin in patients having ocular surgery208, 209. Safe continuation of antithrombotics has been reported for dental210, prostate211 and cardiac surgery212.
The aim of this prospective study was to monitor bleeding complications in a referral based skin cancer centre, and to identify risk factors for this complication.
This study involved patients managed from 1 July 2002 to 28 February 2006 at a skin cancer surgery referral centre in Geelong, Australia. Patients’ antithrombotic medications were not altered either before or following skin cancer surgery unless the INR was over 3.0.
Consecutive patients who were treated during the study interval were included. Surgical procedures included modified Mohs’ micrographic surgery9; small and large excision and closure of lesions; curettage (with or without electrodessication); skin flaps; full and partial thickness skin grafts, and wedge excision surgery. Patients on warfarin therapy must have had an INR test within a week of planned surgery, usually within two days.
Patients who stopped their antithrombotic medication before surgery were excluded, as were patients with an INR above 3.0. Other exclusions were lesions managed entirely by cryotherapy patients who had more than 15 lesions treated (to avoid over representation), and partial and full thickness skin graft donor sites.
One surgeon performed all procedures. All procedures were performed under local anaesthetic. Bipolar diathermy was used to effect haemostasis where appropriate. The site of all removed lesions was recorded and specimens underwent histopathological examination. All full thickness wounds were closed with interrupted nylon skin sutures. Absorbable deep sutures were used only if layers deep to the skin required direct closure. Occlusive wound dressings were used unless patient allergy made this impractical.
In patients with multiple tumours, the main lesion was removed first and a sub analysis of first lesions excised was carried out.
Patients had follow - up at least until removal of sutures. Patients who attended elsewhere for removal of sutures were contacted by telephone.
Post operative bleeding:
Any post operative haemorrhage or haematoma was recorded. A haematoma was regarded as small (up to 5 ml), medium (5 to 50 ml) or large (over 50 ml). A haemorrhage was regarded as small (up to 25 ml), medium (25 to 100 ml), or large (over 100ml).
Haemorrhage was further classified as delayed (1 to 24 hours after surgery) and late (more than 24 hours).
To detect an increased incidence of bleeding of 2.5 per cent compared with 1 per cent, with a sensitivity of 0.05 and a power of 0.8, it was calculated that the study would require 280 procedures on patients taking warfarin. Recruitment continued until at least 280 procedures had been performed on patients taking warfarin and 280 on patients taking aspirin.
Demographic details were presented as percentage or mean (s.d.) as appropriate. Chi-square analysis was used to test the significance of differences between proportions and categorical variables. This method was also used to assess the invariable risk of bleeding, with results presented as odds ratio with 95 per cent confidence interval.(c.i.) Multivariable analysis using binary logistic regression (forward and backward) was used to identify independent risk factors for bleeding, and odds ratio beta-coefficients with 95 per cent were recorded. Receiver - operator characteristic (ROC) curves were used to determine the age cut off value that represented the best combination of sensitivity and specificity for risk of a bleeding complication. SPSS version 14.0.2 (SPSS, Chicago, Illinois, USA) was used for all statistical analysis. P < 0.050 was considered statistically significant.
A total of 5950 skin lesions from 2394 patients were treated by excision or curettage in the 44 months.
Twenty-four patients (29 lesions) were excluded as anticoagulant therapy was stopped before they attended. A further three patients with an INR over 3.0 were excluded. One patient had more than 15 lesions and all lesions after the 15th were excluded from the study.
The mean age of the patients was 64 (17) (median 67) years; with 55.3 per cent were men. The 5950 lesions managed included 3175 malignant lesions; 1436 squamous cell carcinomas (24.1 per cent), 1381 basal cell carcinomas (23.2per cent), 166 melanomas (2.8 per cent), 24 lentigo maligna (0.4 per cent) and 168 other cutaneous malignancies (2.8 per cent). The 2775 benign lesions excised comprised; 1098 benign actinic lesions (18.5 per cent), 137 dysplastic melanocytic naevi (2.3 per cent) and 1540 other benign lesions (25.9 per cent).
Twelve patients were not seen at the time of, or after, removal of sutures and were followed up through telephone contact. All others were seen by the surgeon or nursing staff.
Post operative bleeding
Forty bleeds were recorded (0.7 per cent) among 5950 lesions treated. There were 14 haematomas, ten small and four medium collections. Twenty-six patients had a haemorrhage recorded; there were twenty-two delayed and four late bleeds. Fifteen of the haemorrhages were small, ten were medium and one was large.
The latter occurred 3 weeks after surgery while the patient was on warfarin therapy. Although the INR was therapeutic at the time of surgery, it rose in the following weeks to 7.4. This was the only person admitted to hospital with post operative bleeding. Management involved stopping warfarin and wound compression.
Three patients (two on warfarin) required wound exploration to control bleeding. One patient required vessel ligation and bleeding in the other two was controlled with bipolar diathermy, and compression. Two patients (neither on warfarin) had a haematoma evacuated. Bleeding complications in all other patients were managed conservatively.
Predictors of increased risk of bleeding
There was no difference in the bleed rate between men and women (0.7 versus 0.6 per cent respectively, p= 0.550), but those with bleeding complications were significantly older than the remainder (75 ± 9 versus 64 ± 17 years; p<0.001). ROC curve analysis demonstrated that age 67 years provided the best cut-off value for the risk of a bleeding complication.
There were six bleeds (2.1 per cent) among 285 skin cancer procedures undertaken in 67 patients receiving warfarin therapy only (Table 1). This incidence was significantly higher than in patients not on warfarin) (P<0.001). A total of 829 skin cancer procedures were undertaken on 334 patients taking regular oral aspirin. Nine instances of bleeding were documented (1.1 per cent), and the incidence was higher than in patients who were not on aspirin (p=0.032). There were two bleeds among 35 skin cancer procedures in 11 patients managed with both warfarin and aspirin. The incidence of bleeding was significantly higher than in patients not on antithrombotics, (P < 0.001).
A sub-analysis of the first procedures revealed a significantly increased rate of bleeding in those on warfarin but not those receiving aspirin, (Table 1).
The rate of bleeding after each surgical technique is summarized in Table 2. Only skin grafts were associated with a rate over 4 per cent. Bleeding was least common when wounds were closed directly, (0.4 per cent) or managed by curettage, (0.2 per cent).
Bleeding was not more frequent in diabetics (1.5 per cent; two of 135 first procedures) (P=0.86). Similarly the rate in smokers (0.4 per cent; one of 287) wasn’t different (P=0.491).
Univariable analysis showed that a number of factors were associated with an increased risk of bleeding complications after surgery; age of 67 years or greater, warfarin therapy, aspirin therapy, flap or graft closure, and surgery on or near the ear, (Table 3).
Binary logistic regression revealed that aspirin was not an independent factor as older patients were more likely to take aspirin. The remaining four variables were independent predictors of bleeding complications (Table 3).
Bleeding complications were increased in patients taking warfarin, but not in those taking aspirin. There does not appear to be a case for stopping aspirin before skin procedures. Three other independent risk factors for bleeding were age 67 years or more, surgery in the area of the ear, and flap or graft closure.
The marginal reduction in bleeding incidence risk that might be associated with stopping warfarin therapy cannot be justified uniformly when balanced against the risk of a serious thrombo-embolic event201, 207.
This risk has been variously estimated between one in 278 and one in 11,500193, 202. The variation in these estimates makes it difficult to quantify the proportional risk of minor haemorrhage that a patient would suffer by stopping antithrombotic drugs.
The risk of a major thromboembolic event is estimated at 1 per cent per day if the patient has suffered a deep venous thrombosis or pulmonary embolism within one month of temporarily stopping warfarin cessation.202
Additional precautions to reduce the risk of bleeding may be appropriate when operating on patients taking warfarin. The INR should be at the lower end of therapeutic range, especially when several additional risk factors are present. A fixed INR recommendation cannot be made as the therapeutic range recommended for patients depends on the indication for anticoagulation.202, 204
There was no life threatening bleeds in this study. All were managed with relatively simple measures and bleeding complications experienced were mainly an inconvenience and did not cause significant morbidity. This contrasts with more invasive surgery. Deep cavity and organ bleeding following major surgery may be occult until substantial blood loss has occurred. In particular, abdominal and thoracic cavity bleeding can be life threatening, so the findings of this study cannot be extrapolated to major surgery.
This study has some limitations. No NSAIDS or other anticoagulants were considered. Newer antithrombotics such as ticlopidine201 and clopidogrel were rarely being taken by patients during the design phase of this prospective study, although their usage has increased since. These recommendations regarding aspirin and warfarin cannot be extrapolated to other antithrombotics.
Table 1 Incidence of bleeding complications in patients on warfarin and aspirin
Values in parentheses are percentages. *P<0.001, **P=0.002, ***P=0.003, ****P=0.032 versus patients receiving neither warfarin nor aspirin (x2 test).
Both warfarin & aspirin therapy
Patients on neither warfarin nor aspirin
Number of patients
Total no. skin cancer procedures
Total no. of bleeds
No. of bleeds on first procedures
Table 2: Bleeding incidence following different skin closure techniques
Values in parentheses are percentages. * Versus other closure techniques (x2 test).
Post operative bleeding
Direct primary closure
Skin flap repair
Wedge excision and repair
Table 3: Odds ratios are presented for individual factors found to be associated with bleeding and results of binary logistic regression..
Values in parentheses are 95 per cent confidence intervals. *x2 test.
Binary logistic regression
Age ≥ 67 years
(2.8 , 18.4)
(1.8 , 12.2)
(2.4 , 10.7)
(1.4 , 6.3)
Flap or graft
(2.1 , 7.5)
(1.4 , 5.3)
(1.8 , 8.0)
(1.2 , 5.7)
(1.1 , 4.4)
(0.69 , 2.9)