Bleeding and First-year Mortality Following Hip Fracture Surgery and Preoperative Use of Low-dose Acetylsalicylic Acid

An Observational Cohort Study

Annika M Kragh; Markus Waldén; Anna Apelqvist; Philippe Wagner; Isam Atroshi


BMC Musculoskelet Disord. 2011;12(254) 

In This Article


The original randomized trial has been described in detail elsewhere.[9] The aim of the trial was to evaluate the efficacy if a pneumatic compression bandage applied over the hip after surgery for a hip fracture. Briefly, patients with a proximal femoral fracture presenting at the Emergency Department of Kristianstad Hospital, Sweden, January 2005 through December 2006, were screened by an orthopedic surgeon for enrollment in the trial. The inclusion criteria were patients 50 years or older with cervical fractures planned for hemiarthroplasty or pertrochanteric or subtrochanteric fractures planned for internal fixation with plate and gliding screw or twin hook or with proximal intramedullary nail. The exclusion criteria were non-displaced subcapital (intracapsular) cervical fracture planned for internal fixation, pathologic fracture due to malignancy, concomitant fractures or injuries that might require blood transfusion, and patients refusing blood transfusion. Patients were randomly assigned to the compression group (n = 136) or noncompression group (n = 152) immediately after surgery. The results of the trial showed that the compression bandage did not have any effect on bleeding or blood transfusion requirements. For the present report focusing on preoperative use of LdAA, patients from both arms of the original trial were combined as we did not anticipate that the compression bandage had any bearing on the results.


According to clinical practice prophylaxis against perioperative bleeding was given to all patients with an intravenous injection of tranexamic acid (100 mg/kg body weight) 20 minutes preoperatively and a second injection after 4 hours. Prophylactic treatment against thromboembolic events was given to all patients with a subcutaneous injection of 40 mg enoxaparin daily for at least 10 days after surgery. The transfusion threshold used was blood hemoglobin (Hb) below 100 g/L.

Patients on LdAA (defined as 320 mg daily or lower), dipyridamol or clopidogrel continued with the medication, but patients on warfarin had their medication withdrawn on admission and were managed according to vitamin K antagonist reversing strategy.


At baseline, preoperative use of medication with antithrombotic effect, including warfarin, acetylsalicylic acid or other anti-platelet drugs, and non-steroidal antiinflammatory drugs (NSAIDs), were recorded. At admission, Hb, platelet count, international normalized ratio (INR), activated partial thromboplastin time (APTT), and serum creatinine, were measured. The examining physician obtained medical history and recorded the presence or absence of specific conditions on a preoperative standardized protocol; heart disease (without further specification), hypertension, diabetes, lung disease (specified as asthma, chronic bronchitis or other pulmonary disease), rheumatoid arthritis, and other serious disease. These diagnoses were then verified through review of patients' records by one investigator (AK) who also documented the presence of cardiovascular disease, cerebrovascular disease or both. Before surgery, the anesthesiologist classified the patients according to the American Society of Anesthesiologists (ASA) score.

Intraoperative blood loss was estimated by the nurse anesthetist according to standard procedures. The number of red cell units transfused before, during and after surgery (up to discharge from hospital or death) was recorded. Complications up to 3 months after surgery were recorded. First-year deaths from any cause and the date of death were retrieved from the hospital administrative database linked to the National Board of Health and Welfare's Cause of Death Register.

Sample Size

The pretrial sample size estimation was based on the original trial's two transfusion-related primary outcome measures.[9] As a measure of precision in this analysis we present 95% confidence intervals for the differences in bleeding and transfusion-related variables and mortality.


The study was approved by the Medical Research Ethics Committee of Lund University (704/2004-11-30). Informed consent was obtained from all participating patients or, for patients who could not provide consent themselves due to cognitive impairment, from a family member.

Statistical Analysis

Baseline characteristics of the patients using and those not using LdAA preoperatively were compared with the t-test for continuous variables and Fisher's exact test for proportions. Bleeding and transfusion-related variables were compared between the two groups using logistic regression or analysis of covariance (ANCOVA) adjusting for age (as continuous variable), sex, baseline Hb and type of surgery and odds ratios and 95% confidence intervals (CI) were calculated. A Kaplan-Meier survival curve was constructed to compare first-year mortality among patients according to preoperative use of LdAA. The Fisher exact test was used to compare mortality according to patient sex and dose of preoperative LdAA. The Fisher exact test was also used to compare postoperative complications in LdAA users and non-users. A Cox regression analysis was performed with first-year mortality as the dependent variable. The independent variables were age, sex, LdAA at the time of fracture, type of fracture, baseline cardiovascular and/or cerebrovascular disease and renal dysfunction. The hazard ratio (HR) with 95% CI was calculated. Because patients with hypertension were classified as having a cardiovascular disease and the ASA score (I, II or III/IV) was essentially based on cardiovascular disease we did not include these two variables in the model. To investigate whether a relationship between preoperative use of LdAA and mortality may differ between women and men we repeated the Cox regression analysis adding a term for the interaction between sex and preoperative LdAA use to the model. Adding compression use as a covariate in the models did not change the results. All analyses were two-sided and a P value < 0.05 was considered to indicate statistical significance.