Timing of heparin prophylaxis and bleeding complications in hysterectomy a nationwide prospective cohort study of 9,949 Danish women

Transcription

1 Acta Obstetricia et Gynecologica. 2008; 87: ORIGINAL ARTICLE Timing of heparin prophylaxis and bleeding complications in hysterectomy a nationwide prospective cohort study of 9,949 Danish women CHARLOTTE T. HANSEN 1, HENRIK KEHLET 2, CHARLOTTE MØLLER 3, LINA MØRCH 4, JAN UTZON 5 & BENT OTTESEN 6 1 Research Centre for Prevention and Health, 84/85, Glostrup University Hospital, Ndr. Ringvej 57, Glostrup, 2600 Denmark, 2 Rigshospitalet, Section of Surgical Pathophysiology, Copenhagen, Denmark, 3 Department of Obstetrics and Gynecology, Skejby University Hospital, Brendstrupgaardsvej 100, Aarhus, 8200 Denmark, 4 Research Centre for Prevention and Health, Glostrup Hospital, Nordre Ringvej 57, Glostrup, 2600 Denmark, 5 Unit for Clinical Quality, Bispebjerg Hospital, Bispebjerg Bakke 23, Copenhagen, 2400 Denmark, and 6 Juliane Marie Centre, University Hospital of Copenhagen, Rigshospitalet, Blegdamsvej 9, Copenhagen, 2100 Denmark Abstract Objective. To examine bleeding complications and thromboembolic events in relation to timing of heparin prophylaxis after hysterectomy. Design. Nationwide prospective cohort study with 30 days post-operative follow-up within the Danish Hysterectomy Database (DHD). Setting. All gynecological departments in Denmark (n31). Sample. 9,949 women who had an elective hysterectomy for benign indication between October 2003 and May 2006 and were reported to DHD (national response rate: 8899% throughout ). Main outcome measures. Odds ratios (OR) of peri-operative bleeding complications (]1,000 ml bleeding during surgery or post-operative wound/vaginal-vault/intraabdominal bleeding or hematoma) and number of events of venous thromboembolism. Logistic regression analysis adjusting for: age, body mass index, alcohol, smoking, meno-/metrorrhagia, uterine weight, department volume, surgeon s experience, route and type of hysterectomy and additional surgery, and stratification on assistant s experience, peri-operative pain prophylaxis with NSAID and daily use of Acetyl Salicylic Acid (ASA)/NSAID. Results. 9,051 women (92%) received thromboprophylaxis with heparin, initiated pre-operatively in 48% and post-operatively in 52%. At least one bleeding complication was noted in 881 women (10%). Post-operative heparin administration was associated with a reduced risk of bleeding complications; OR0.85 (95% confidence interval ) compared to pre-operative administration. Excluding cases with potential impaired hemostasis at baseline, the OR was 0.78 ( ). There was no fatal embolism. Three of seven pulmonary embolisms and one of three symptomatic deep venous thromboses occurred with the post-operative heparin administration. Conclusion. Post-operative rather than pre-operative administration of heparin prophylaxis may reduce the risk of bleeding complications after hysterectomy without apparent risk of increased thromboembolic events. Key words: Hysterectomy, heparin timing, bleeding complications, thromboembolisms, national database Introduction Peri-operative bleeding complications remain a significant problem after hysterectomy, despite use of minimally invasive techniques, selective cyclooxygenase-2 inhibitors and bridging therapy with short acting anti-coagulants. Thromboembolic prophylaxis is indicated for major surgery in high risk patients, including patients undergoing hysterectomy (1). In this context, timing of thromboembolic prophylaxis is essential for efficacy and safety (2) and studies from major orthopedic procedures suggest that post-operative initiation of thromboembolic prophylaxis may be as effective as pre-operative administration but with potential less bleeding complications (26). Correspondence: Charlotte Thim Hansen, Research Centre for Prevention and Health, 84/85, Glostrup University Hospital, Ndr. Ringvej 57, Glostrup, 2600 Denmark. thim@dadlnet.dk (Received 18 July 2008; accepted 7 August 2008) ISSN print/issn online # 2008 Informa UK Ltd. (Informa Healthcare, Taylor & Francis AS) DOI: /

2 1040 C. T. Hansen et al. No international consensus on timing exists in surgical gynecology (7), and today prophylaxis is initiated either before or after surgery in hysterectomy (8). Facing a nationwide problem in Denmark with 6% bleeding complications after simple hysterectomy (9), we wanted to study the clinical impact of the pre-operative vs. post-operative heparin administration, focusing on both bleeding complications and thromboembolism. Material and methods We studied all women reported to the Danish Hysterectomy Database (DHD) between October 2003 and May 2006 inclusive. In this period the database enrolled 9,949 women undergoing elective hysterectomy for benign disease, (women with malignant disease or acute hysterectomy, for example, post-partum hysterectomy were excluded in DHD). The DHD collaboration was established in 2003 to monitor and improve surgical quality of benign hysterectomy in Denmark. The clinical database opened for registration on October 2003, and since June 15, 2006 the registration has been obligatory by law (announcement nr. 459 of May 16, 2006 National Board of Health, Denmark). The national response rate for DHD was 88% in the first half of 2004, 90% in the second half of 2004 and 99% throughout 2005, when the Danish National Patient Registry (LPR) was used as gold standard (personal correspondence). LPR contains information on all patient contact with clinical hospital departments, including re-admissions at gynecological departments or elsewhere in Denmark (10). The data collection process in DHD were based on a structured registration form, which was completed prospectively by the surgeon whenever a woman was referred for a hysterectomy on a benign indication. The registration form consists of three parts; Page 1 includes baseline information about lifestyle, disease and indication for surgery (completed before surgery), Page 2 collects information about surgical procedures, possible prophylactic regimens and the surgeons experience (completed immediately following surgery) and Page 3 has information about post-operative hospitalization, complications and re-operations (completed at discharge). If the patient was re-admitted at a gynecological department, a separate registration form about complications and re-operations was completed at discharge and reported to LPR. If the patient was transferred or re-admitted to another hospital department, outcome was routinely reported to LPR by the hospital department admitting the patient. Data were submitted to LPR without using the structured data template from DHD, but in accordance with a declared national standard including information about the unique personal identification number (given to each citizen in Denmark), the diagnosis and the performed procedures. The follow-up period was 30 days postoperatively based upon data from LPR. The data were reported online to LPR by the departments, using the unique personal identification number (CPR number) for each citizen in Denmark, to link the information. Subsequently, the registration form was mailed to the central DHD office for manual validation. Incomplete (or incorrectly completed) registration forms were returned to the departments for correction and new online registration in LPR. Data on mortality within 30 days post-operatively were available from the Danish Civil Registration System (CRS), and the cause of death was retrieved from death certificates. In accordance with the study protocol, we excluded women with known coagulopathies or daily anticoagulant therapy with K-vitamin antagonists at baseline (n 123). During some gynecological departments in Denmark recommended the pre-operative heparin administration, some recommended postoperative administration and in a few departments it was the surgeon who made the final decision. The heparin was unfractionated (UFH) or lowmolecular-weight heparin (LMWH) according to local instruction or by surgeon s preference and was administered daily until discharge. Timing was registered by the surgeon, and categorized as preoperatively if the first dose of heparin was administered before incision and as postoperatively if the first dose of heparin was initiated after surgery. The primary outcome variable was the number of women with one or more bleeding complications, within one month post-operatively: (1) ]1,000 ml bleeding during surgery; or (2) post-operative vagi nal-vault/wound/intraabdominal bleeding or hematoma. The surgeon registered the amount of bleeding (in milliliters) during surgery. For analysis we made a cutoff at 1,000 ml bleeding or more. By the time of discharge from primary hospitalization or readmission(s) the physician reported information about possible post-operative bleeding complications. According to the definition on the registration form only post-operative bleeding complications causing re-operation or prolonged hospitalization (exceeding local instructions) were registered in DHD.

3 Heparin timing in hysterectomy 1041 The outcome variable, with respect to efficacy, was the number of women with one or more thromboembolic events (i.e. deep vein thrombosis (DVT) or pulmonary embolism) within one month post-operatively, registered during primary hospitalization or re-admission(s). Only clinically evident thromboembolism was registered in DHD. Because pre- vs. post-operative heparin administration was not assigned randomly to the study subjects the two groups are likely to differ on a number of variables that might confound or distort the causal effect of heparin timing. The age of the women, at the day of surgery, was obtained from the CPR which includes the date of birth. Height and weight was self-reported. Body mass index (BMI) was calculated (kg/m 2 ). Alcohol consumption was reported as current average weekly intake of alcohol (one unit equals one glass of wine (12.5 cl), one bottle of beer (33 cl) or one glass of aquavit/spirit (4 cl). We categorized alcohol consumption as: (I) no unit of alcohol per week; (II) 17 units per week; (III) 814 units per week; or (IV) 14 units per week. Tobacco use was reported as current daily quantity of tobacco smoked (one cigarette was equivalent to 1 g, a cheroot to 3 g and a cigar to 5 g of tobacco). We made a binary variable for smoking status: Non-smoker: 0 g tobacco per day or Smoker: 0 g tobacco per day. Women were categorized as having an abnormal uterine bleeding if menorrhagia or metrorrhagia was among the indications for surgery reported to DHD. Uterine weight (g) was measured on an electronic weight in the operating theatre immediately after removal of the uterus. The department specific volume was obtained from LPR and is defined as the total number of benign hysterectomies performed within each department in The annual volume was assumed to be a proxy for the department effect. The experience of the surgeon and the assistant was assessed as the total number of previously performed hysterectomies of the same type (abdominal, vaginal or laparoscopically assisted). Both variables were reported by the surgeon in the following categories; 510, 1125, 2650, 5175, 76100, 100 operations performed. All procedures were reported by the surgeon. The type of hysterectomy was categorized as total (including the cervix) or subtotal, and the surgical approach as abdominal, vaginal or laparoscopically assisted hysterectomy (8,11). Additional surgery to the hysterectomy was categorized as a binary variable: yes, for adnexal surgery, considerable adhesiolysis or colporrhaphy and no, if hysterectomy was the only procedure. Women with daily use of antiplatelet drugs as aspirin (ASA) or non-steroidal anti-inflammatory drugs (NSAID) were categorized as Daily ASA/ NSAID-user if this medication was not stopped in time before surgery (i.e. seven and three days preoperatively, respectively (12,13)) and as No daily ASA/NSAID-user if the medication was stopped in time before surgery or the women had not been on daily treatment with ASA or NSAID. A single dose of pain prophylaxis with NSAID, prescribed by the surgeon pre- or intra-operatively, was categorized as a binary variable (yes/no), excluding use of selective cyclooxygenase-2 (COX-2) inhibitors, as they do not influence platelet aggregation (14,15). Statistical analysis The association between heparin timing and the risk of bleeding complications was analyzed by logistic regression model estimating odds ratios (OR) with 95% confidence intervals (CI). Pre-operative heparin administration was chosen as reference and the Score test was used for significance testing (p B0.05). Potential confounder variables were chosen a priori based on existing evidence or biological plausibility. Firstly, we adjusted for age. Secondly, we adjusted for all the assumed confounder variables ( full model ): age, BMI, meno-/metrorrhagia, uterine weight, department volume, surgeons experience, route and type of hysterectomy and additional surgery. Information on smoking status and alcohol consumption was available for 87 and 84%, respectively. Thus, we adjusted for smoking and alcohol within subgroups of the total study population. The continuous variables were modeled as linear or as second-degree polynomials, where appropriate. We assumed that the experience of the assistant could modify the association between the surgeons experience and bleeding complications, and analyzed this by nesting in a separate subgroup. To assertain that daily antiplatelet drugs (ASA/ NSAID) did not confound or modify the investigated association between heparin timing and bleeding risk we examined the effect of Daily use of ASA/ NSAID and pre- and intra-operative pain prophylaxis with NSAID by stratification. Subsequently, we included an interaction term to test for effect modification (p B0.05). Similarly, the effect of smoking and alcohol consumption was examined. Women with missing information on values used in the models were excluded. Model fit was evaluated by the Hosmer and Lemeshow test (16). Statistical software was SAS version 9.1.

4 1042 C. T. Hansen et al. Results The majority (92%) received thromboprophylaxis with heparin (n 9,051). Complete information about timing of the first dose of heparin was available in 99% (n 9,035). The first dose of heparin was initiated pre-operatively in 48% and post-operatively in 52% of the women (Figure 1). Complete information about one or more bleeding complications was available for 8,980 women (99%) of which 881 women (10%) had at least one bleeding complication. Table I shows the characteristics of the study population in relation to heparin timing. In general baseline differences between groups were small. Compared to women with pre-operative heparin administration, women with post-operative heparin administration more often had a total hysterectomy, had less often an abdominal hysterectomy and were less often operated upon by a highly experienced surgeon, but more often had an experienced surgeon as assistant surgeon. Two percent of all hysterectomies, in both groups, were done without an assistant surgeon. Compared to women with preoperative heparin administration, women with postoperative heparin administration had a higher department specific volume, a smaller median uterine weight and were less frequently exposed to daily NSAIDs or ASA not stopped in time before surgery, but more frequently used pain prophylaxis with NSAIDs immediately before or intra-operatively. The crude distribution showed significantly less bleeding complications in the post-operative heparin group compared to the pre-operative heparin group (9% vs. 11%, respectively; p 0.005). Fewer women with excessive intra-operative bleeding (]1,000ml) were observed in the post-operative heparin group compared to the pre-operative heparin group (2.5% vs. 3.2%, respectively; p 0.04) and fewer women with post-operative bleeding complications were observed in the post-operative heparin group compared to the pre-operative heparin group (7% vs. 8%, respectively; p 0.002). Adjustment for age showed an 18% lower risk of bleeding complications with the post-operative heparin administration compared to pre-operative heparin administration (n 8,980) (Table II). The final multivariate adjusted analysis was performed on women with complete information on heparin timing, bleeding complications and all confounding factors shown in Table I (n 8,001). Again, post-operative heparin administration was associated with the lowest risk of bleeding, with a 15% reduction in risk of bleeding complications, compared to the pre-operative initiation (Table II). To check for selection bias, we made a crude age adjusted analysis on the same reduced dataset as used in the multivariate adjusted analysis (n8,001) and found no change in the point estimate (OR 0.83, CI ). In the final model, additional adjustments for alcohol consumption (n6,860) and smoking status (n7,097) did not hysterectomy 9,826 included 123 excluded coagulopathies or daily anticoagulant therapy * 2 missing information heparin prophylaxis 9,051 heparin prophylaxis 773 no heparin prophylaxis 16 missing information heparin timing 4,348 preoperative heparin analysed 4,687 postoperative heparin analysed Figure 1. Study population. Between October 2003 and May 2006 inclusive the Danish Hysterectomy Database enrolled 9,949 women. In this study we excluded 914 women (123 with known coagulopathies or daily anticoagulant therapy with K-vitamin antagonists, 773 without heparin prophylaxis and 18 women with missing information about heparin prophylaxis or timing) leaving 9,035 women for the final analysis.

5 Heparin timing in hysterectomy 1043 Table I. Baseline characteristics and prognostic factors in relation to timing of heparin prophylaxis among women (n9,035) with available information on heparin timing (99%) reported to the Danish Hysterectomy Database between October 2003 and May Pre-operative heparin n4,348 Post-operative heparin n4,687 Missing p-value Patient Age (years)* 47 (4353) 47 (4354) 0 0.2** Body mass index (Kg/m 2 )* 25 (2228) 25 (2228) ** Body mass index (Kg/m 2 ) n (%) $ B (1) 85 (2) 18.5B25 2,078 (49) 2,283 (50) 25B30 (overweight) 1,375 (33) 1,452 (32) ]30 (obesity) 712 (17) 780 (17) Smokers n (%) 1,290 (35) 1,401 (34) 1, $ Alcohol consumed (unit/week) n (%) 1, $ 0 2,102 (59) 2,310 (57) 17 1,103 (31) 1,326 (33) (8) 315 (8) 14 % 70 (2) 95 (2) Uterus weight (g)* 177 (100385) 161 (93363) ** Uterus weight (g) n (%) $ ,539 (66) 2,989 (69) (13) 498 (11) 4501, (16) 693 (16) 1, (5) 179 (4) Menorrhagia/metrorrhagia as indication n (%) 2,239 (52) 2,397 (51) $ Process Surgical approach n (%) 1 B $ Abdominal 2,844 (65) 2,797 (60) Vaginal 1,270 (29) 1,608 (34) Laparoscopically assisted 234 (5) 281 (6) Type of hysterectomy n (%) 7 B $ Total 3,811 (88) 4,234 (90) Subtotal 537 (12) 446 (10) Additional surgery n (%) 1,968 (45) 2,187 (47) $ Daily ASA/NSAID not paused 7/3 days pre-op n (%) 299 (7) 266 (6) $ No daily use of ASA/NSAID 3,984 (92) 4,303 (92) Daily use of ASA/NSAID, but paused 64 (1) 114 (2) Daily use of ASA/NSAID, not paused 299 (7) 266 (6) Pain prophylaxis (NSAID) pre/peri-op 1,215 (28) 1,794 (38) 7 B $ Organization Surgeon s experience (total volume of same $ procedure) n (%) (8) 418 (9) (11) 500 (11) (14) 663 (14) (7) 417 (9) (7) 349 (7) 100 2,291 (53) 2,335 (50) Assistant s experience (total volume of same 96 B $ procedure) n (%) 510 1,976 (46) 1,911 (41) (11) 494 (11) (6) 295 (6) (3) 167 (4) (3) 151 (3) 100 1,279 (30) 1,535 (33) No assistant present 82 (2) 91 (2) Annual volume of hysterectomies/department* 177 (152220) 185 (152219) B0.0001** n Number of women. *Median (IQR). **Wilcoxon signed rank sum. $ Chi square test. % In the group with a weekly alcohol intake above national sensible drinking limits for women (i.e. 14 unit/week) the median intake (25th and 75th percentiles) was similar in the two groups: 20 (16,21) in the pre-operative group and 20 (18,21) in the post-operative group.

6 1044 C. T. Hansen et al. change the point estimate (OR 0.84, CI and OR0.83, CI ), and no effect modification was found. Allowing for effect modification of surgeon s experience by the assistant s experience (by nesting) showed no change in point estimate (OR 0.84, CI ). Pain prophylaxis with NSAIDs prescribed as a single dose, pre- or intra-operatively, was used in one-third of the women being hysterectomized, but we found no change in point estimate (OR0.84, CI ) and no significant effect modification (p 0.70). Eight percent (n 743) of the women received daily treatment with ASA or NSAIDs before surgery, and only one-fourth (n 178) had treatment cessation before surgery. Among the women with no daily ASA or NSAID treatment, or who had the treatment stopped in time before surgery, post-operative heparin administration was associated with a reduced risk of bleeding complications (OR 0.82, CI ) compared to pre-operative heparin administration. In contrast, among the women with daily use of ASA or NSAIDs not stopped in time before surgery, postoperative heparin administration was associated with an increased risk of bleeding complications (OR 1.55, CI ) compared to the pre-operative heparin administration. Daily use of ASA or NSAIDs, not stopped in time before surgery, significantly modified the effect of heparin timing on the risk of bleeding complications (p 0.04). In a post hoc analysis, excluding all cases with potential impaired hemostasis (i.e. women with NSAID pain prophylaxis before or during surgery or women on daily treatment with ASA/NSAID not stopped in time before surgery), the beneficial effect of post- vs. pre-operative heparin showed a reduced OR of 0.78 (CI ) (n 5,003) and an estimated reduction in the risk of bleeding complications of 22%. Mechanical thromboprophylaxis with graduated elastic compression stockings was used less often among women with post-operative heparin initiation, compared to women with pre-operative heparin initiation (n 2,869 (61%) vs. n 3,606 (83%); pb0.0001). There was no significant difference in the number of thromboembolic events according to heparin timing. Deep venous thrombosis was observed in one vs. two women (p0.5) and pulmonary embolism in three vs. four women (p0.6) among the women with post-operative vs. pre-operative heparin administration. No death was observed within one month post-operatively. The small number of venous thromboembolism cases precluded logistic regression modelling. The median hospitalization was two days (25%: two days, 75%: three days) in both groups. Discussion We found that a significant reduction of perioperative bleeding complications was associated with postponing the first dose of heparin thromboprophylaxis until after surgery, based upon prospectively collected nationwide data from the DHD. Thus, the post-operative heparin regimen was associated with a 1522% reduction in the risk of bleeding complications compared to the pre-operative heparin administration. Importantly, post-operative initiation was not associated with any increase in the otherwise very low number of thromboembolic events. Our results are consistent with a previous randomized study in hip surgery where LMWH administered around six hours post-operatively was effect neutral to pre-operative administration regarding thromboembolism, but where pre-operative LMWH was associated with a significant increase in major bleeding complications (3). In our study 3% of the women lost more than 1,000 ml of blood intra-operatively and 7% had one or more postoperative bleeding complications, which was at the same level as seen in national (9,17) and international reference studies (1820). In the absence of thromboprophylaxis the incidence of DVT ranges from 6 to 29% among patients undergoing major gynecologic operations for benign disease (21). The small number of DVT (0.03%) and embolism (0.08%) observed in our study precluded adjustment for potential confounding factors. However, it must be emphasized that these numbers reflect symptomatic events, which are much fewer than those assessed by more sensitive methods in randomized trials (RCT) or has been observed in observational studies with infrequent use of medical thromboprophylaxis (22). In this study we examined if daily antiplatelet drugs (ASA/NSAID) confounded or modified the observed association between heparin timing and bleeding risk. Surprisingly, women on daily ASA or NSAIDs (n 565) had an increased risk of bleeding when heparin was administered post-operatively compared to pre-operative administration. However, the interpretation of this effect should be cautious, because this estimate is based on a small number of women. To clarify the effect of heparin timing

7 Heparin timing in hysterectomy 1045 Table II. Association between heparin timing and bleeding complications (]1,000 ml bleeding during surgery or post-operative vaginalvault/wound/intraabdominal bleeding or haematoma) in women (n9,035) with benign hysterectomy performed between October 2003 and May 2006 and reported to the Danish Hysterectomy Database. Bleeding complications Age adjusted Multivariate adjusted* Heparin timing (First dose administered) Number of women $ Number cases OR (95% CI) p-value Number of women $ Number cases OR (95% CI) p-value Pre-operatively 8, , Post-operatively ( ) ( ) *Adjusted for age (linear), body mass index (linear and squared), uterus weight (linear and squared), menorrhagia /metrorrhagia (yes or no), department specific volume (linear), surgeon s experience (510, 1125, 2650, 5175, 76100, ]100 total operations of same procedure), surgical approach (abdominal, vaginal or laparoscopically assisted), type of hysterectomy (total or subtotal), additional surgery (yes or no). $ Women with complete information on all variables used in the adjusted analysis (55 women not included because of missing information on bleeding complications). CI, Confidence interval. among women on daily ASA/NSAIDs further investigation is needed. Excluding women who have been exposed to ASA or NSAIDs (either as peri-operative pain prophylaxis or as daily pre-operative treatment) a more pronounced beneficial effect was observed from post-operative heparin administration with a 22% reduction in the risk for peri-operative bleeding complications, compared to the pre-operative heparin administration. Interestingly, pain prophylaxis with NSAIDs prescribed as a single-dose immediately before or intra-operatively did not modify the positive effect of post- vs. pre-operative heparin administration on bleeding complications. The strength of this study is the large sample size (almost 10,000 women) and nationwide complete follow-up data on almost all women undergoing hysterectomy in Denmark in the 2.5 year period studied. Furthermore, completeness of data is high (91100%), with the exception of information about alcohol consumption (84%) and smoking habits (87%). Adjustment for alcohol and smoking did not change our results because missing information on alcohol and smoking habits are thought to be random, i.e. due to inappropriate organization of the data collection rather than because of a deviant status of the drinking or smoking variable, the stratified analysis is believed to be valid. In our study, the median hospitalization was two days due to intensive national focus on fast-trackregimen in hysterectomy (23). This multimodal approach includes thorough pre-operative information, multimodal pain treatment and early mobilization and discharge. Early mobilization might in itself reduce the risk of thromboembolic events and therefore potentially reduce the generalizability of the results of delayed heparin administration with longer stays. A limitation of our study is also the nonrandomized design with risk of unmeasured confounding. However, a randomized controlled trial to address the present topic would have important methodological limitations. To evaluate both safety and efficacy of a large sample size, equal to the size of our cohort, it would be necessary to study rare, but potentially fatal, thromboembolic events. Due to low recruitment rates observed in gynecological surgery (17) a large randomized trial would be very time- and cost-consuming and likely to be biased from time trends in surgical indications or performance. Nevertheless, our extensive statistical adjustment for age, BMI, menorrhagia/metrorrhagia, uterine weight, surgical approach, type of hysterectomy, surgeons experience and department specific annual volume, does allow controlling for

8 1046 C. T. Hansen et al. prognostically important baseline differences among the groups, including residual confounding correlated with choice of surgical procedure, and supporting the reliability of our results. Consequently, the reported ORs cannot be confounded by uterine weight, surgical approach or other prognostic conditions correlated with choice of abdominal hysterectomy. Information about type (i.e. UFH vs. LMWH) and dosage of heparin was not available in our cohort. However, in 2006, a D anish questionnaire survey examined the use of heparin prophylaxis in gynecological surgery. Local instructions were returned from 20 out of 31 available gynecological departments in Denmark. In all instructions LMWH was recommended (personal communication). Therefore, it is most unlikely that heparin type could have biased the effect of heparin timing in our study. Additionally, conflicting results have been found concerning safety against bleeding complications (24,25), although LMWH is more convenient with fewer daily doses required (24). In our study, the assessors of bleeding complications were not blinded to heparin exposure. However, possible misclassification was most likely nondifferential, because potential underestimation of outcome was presumably similar in both groups as the choice of heparin timing depended on local beliefs. In addition, the objective of running this clinical database was overall quality surveillance and improvement. Thus, a random error of this kind will not distort our results, but only underestimate the true beneficial effect associated with the post-operative heparin administration (26,27). National audits, within the framework of the DHD, have identified a great variety in timing of the first dose of heparin. In Denmark the postoperative regimen was initiated between 2 and 15 hours after surgery. In hip surgery, early postoperative administration (Bfour hours post-operatively) was associated with an increased bleeding risk (6). This might have influenced our results, as we did not distinguish between early and later postoperative administration. Therefore, inclusion of women with early post-operative heparin administration might have underestimated the advantage of the post-operative heparin administration. In elective hip arthroplasty the peak efficacy for LMWH against thromboembolism ranges between two hours pre-operatively and 68 hours post-operatively (5). The future challenge is to identify the optimal time window for the first post-operative heparin dose, regarding safety and efficiency after hysterectomy. In conclusion, a post-operative anti-thrombotic heparin regimen reduced bleeding complications but preserved anti-thrombotic efficacy after hysterectomy. The Danish hysterectomy database (DHD) collaboration The Steering committee: Jan Blaakær, MD, DMSc, Margit Dueholm, MD, PHD., Eriksen Gitte, MD, PHD., Helga Gimbel, MD, DMSc, Charlotte T Hansen, MD, Peter Hornnes, MD, DMSc, Henrik Kehlet, MD, DMSc, Niels Kjærgaard, MD, PHD., Ulla Breth Knudsen, MD, PHD., Lydia De Lassen, MD, PHD., Øjvind Lidegaard, MD, DMSc, Ole Mogensen, MD, DMSc, Charlotte Møller, MD, PHD, Dorte Nielsen, MD, Bent Ottesen, MD, DMSc, Kjeld Leisgaard Rasmussen, MD$, Annette Settnes, MD, PHD., Jan Utzon, MD. Participating Gynecological Departments Bornholms Hospital, Esbjerg Centralsygehus, Fredericia Hospital, Frederiksberg Hospital, Frederikshavn Hospital, Gentofte University Hospital, Glostrup University, Grindsted Hospital, Haderslev Hospital, Herlev University, Herning Centralsygehus, Hillerød Hospital, Hjørring Hospital, Holbæk Centralsygehus, Horsens Hospital, H:S Hvidovre University Hospital, H:S Rigshospitalet University Hospital, Kolding Sygehus, Nykøbing F Centralsygehus, Næstved Centralsygehus, Odense University Hospital, Randers Centralsygehus, Roskilde Amtssygehus, Silkeborg Centralsygehus, Skejby University Hospital, Slagelse Centralsygehus, Svendborg Hospital, Sønderborg Hospital, Thisted Hospital, Viborg Hospital, Aalborg Hospital. Funding The Danish Hysterectomy Database was supported by the Danish county council, which had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The authors had full access to all the data in the study and had final responsibility for the decision to submit for publication. Ethical approval DHD is registered according to the Danish law of data protection (Datatilsynet). No ethics review board approval or patient s consent is required because the database is accredited by the Danish National Board of Health.

9 Heparin timing in hysterectomy 1047 Acknowledgements Statistician Steen Ladelund, Research Centre for Prevention and Health, Glostrup University Hospital, is acknowledged for substantial statistical guidance. References 1. Geerts WH, Pineo GF, Heit JA, Bergqvist D, Lassen MR, Colwell CW, et al. Prevention of venous thromboembolism: the Seventh ACCP conference on antithrombotic and thrombolytic therapy. Chest. 2004;/126:/ Raskob GE, Hirsh J. Controversies in timing of the first dose of anticoagulant prophylaxis against venous thromboembolism after major orthopedic surgery. Chest. 2003;/124:/ Hull RD, Pineo GF, Francis C, Bergqvist D, Fellenius C, Soderberg K, et al. Low-molecular-weight heparin prophylaxis using Dalteparin in close proximity to surgery vs. Warfarin in hip arthroplasty patients: a double-blind, randomised comparison: North American Fragmin Trial Investigators. Arch Intern Med. 2000;/160:/ Hull RD, Brant RF, Pineo GF, Stein PD, Raskob GE, Valentine KA. Preoperative vs. postoperative initiation of low-molecular-weight heparin prophylaxis against venous thromboembolism in patients undergoing elective hip replacement. Arch Intern Med. 1999;/156:/ Hull RD, Pineo GF, Stein PD, Mah AF, MacIsaac SM, Dahl OE, et al. Timing of initial administration of low-molecularweight heparin prophylaxis against deep vein thrombosis in patients following elective hip arthroplasty. Arch Intern Med. 2001;/161:/ Strebel N, Prins M, Agnelli G, Büller HR. Preoperative or postoperative start of prophylaxis for venous thromboembolism with low-molecular-weight heparin in elective hip surgery? Arch Intern Med. 2002;/162:/ Harris WJ. Complications of hysterectomy. Clin Obstet Gynecol. 1997;/40:/ Johnson N, Barlow D, Lethaby A, Tavender E, Curr E, Garry R. Surgical approach to hysterectomy for benign gynaecological disease. Cochrane Database of Systematic Reviews 2006, Issue 2. Art. No.: CD DOI: / CD pub3. 9. Moller C, Kehlet H, Utzon J, Ottesen B. Hysterectomy in Denmark. An analysis of postoperative hospitalisation, morbidity and readmission. Dan Med Bull. 2002;/49:/ Andersen TF, Madsen M, Jørgensen J, Mellemkjoer L, Olsen JH. The Danish National Hospital Register. A valuable source of data for modern health sciences. Dan Med Bull. 1999;/46:/ Lethaby A, Ivanova V, Johnson NP. Total versus subtotal hysterectomy for benign gynaecological conditions. Cochrane Database of Systematic Reviews 2006, Issue 2. Art. No.: CD DOI: / CD pub Mueller RL. First-generation agents: aspirin, heparin and coumarins. Best Pract Res Clin Haematol. 2004;/17:/ Slappendel R, Weber EW, Benraad B, Dirksen R, Bugter ML. Does ibuprofen increase perioperative blood loss during hip arthroplasty? Eur J Anaesthesiol. 2002;/19:/ Hegi TR, Bombeli T, Seifert B, Baumann PC, Haller U, Zalunardo MP. Effect of rofecoxib on platelet aggregation and blood loss in gynaecological and breast surgery compared with diclofenac. Br J Anaesth. 2004;/92:/ Zemmel MH. The role of COX-2 inhibitors in the perioperative setting: efficacy and safety a systematic review. Am Ass Nur Anesth J. 2006;/74:/ Hosmer DW, Lemeshow S. Applied logistic regression. USA: John Wiley & Sons, Inc; Gimbel H, Zobbe V, Andersen BM, Filtenborg T, Gluud C, Tabor A. Randomised controlled trial of total compared with subtotal hysterectomy with one-year follow up results. BJOG. 2003;/110:/ Beckmann M, Neppe C. Morbidity associated with vaginal hysterectomies in Queensland public teaching hospitals. Aust NZ J Obstet Gynaecol. 2007;/47:/ Dicker R, Greenspan JR, Strauss LT, Cowart MR, Scally MJ, Peterson HB, et al. Complications of abdominal and vaginal hysterectomy among women of reproductive age in the United States. Am J Obstet Gynecol. 1982;/144:/ Thakar R, Ayers S, Clarkson P, Stanton S, Manyonda I. Outcomes after total versus subtotal abdominal hysterectomy. N Engl J Med. 2002;/347:/ Davis JD. Prevention, diagnosis, and treatment of venous thromboembolic complications of gynaecologic surgery. Am J Obstet Gynecol. 2001;/184:/ Mäkinen J, Johansson J, Tomás C, Tomás E, Heinonen PK, Laatikainen T, et al. Morbidity of hysterectomies by type of approach. Hum Reprod. 2001;/16:/ Møller C, Kehlet H, Friland SG, Schouenborg LO, Lund C, Ottesen B. Fast track hysterectomy. Eur J Obstet Gynaecol Reprod Biol. 2001;/98:/ Palmer AJ, Schramm W, Kirchhof B, Bergmann R. Low molecular weight heparin and unfractionated heparin for prevention of thrombo-embolism in general surgery: a metaanalysis of randomised clinical trials. Haemostasis. 1997;/27:/ Cook G, Depares J, Singh M, McElduff P. Readmission after hysterectomy and prophylactic low molecular weight heparin: retrospective case-control study. BMJ. 2006;/332:/ Sorensen HT, Lash TL, Rothman KJ. Beyond randomized controlled trials: a critical comparison of trials with nonrandomized studies. Hepatology. 2006;/44:/ Grimes DA, Schulz KF. Bias and causal associations in observational research. Lancet. 2002;/359:/24852.