Study Design
The Danish National Health Service provides tax-supported health care for all Danish citizens, guaranteeing free medical care for both emergency and general admissions to hospitals and outpatient clinics. Denmark has a total population of approximately 5.5 million inhabitants.
Our study used a population-based follow-up design, which was based on data from prospective, nationwide Danish medical registries. The Danish Hip Arthroplasty Register25 contains data on primary total hip replacements and revisions performed in Denmark since 1995. All forty-five orthopaedic departments in Denmark, including six private hospitals, provide data to this registry. Preoperative and perioperative data are recorded by the operating surgeon.
The Danish National Registry of Patients has maintained a database of all admissions and discharges and up to twenty diagnoses for each admission to public nonpsychiatric hospitals in Denmark since 1977. Diagnoses are classified according to the International Classification of Diseases (ICD). The eighth edition (ICD-8) was used from 1977 to 1993, and the tenth edition (ICD-10) has been used since 1993. All diagnoses are coded by the physician discharging the patient.
The Danish Civil Registration System has maintained a database of information on vital status, date of death, residence, and migration for the entire Danish population since 1968. All Danish citizens are assigned a unique ten-digit personal identification number at birth or on immigration, permitting unambiguous linkage among all Danish administrative registries, as well as tracking of patients who died, emigrated, or were admitted to different hospitals26.
The Danish Registry of Causes of Death maintains a repository of death certificates for all citizens who died in Denmark since 1970. Cause of death is coded according to the ICD by the physician with the most accurate knowledge of the events leading to death, with one underlying cause of death and up to three additional immediate causes.
Patients
From the Danish Hip Arthroplasty Register, we identified 68,863 primary total hip replacement procedures performed on 58,669 patients from January 1, 1995, to December 31, 2006. All patients received pharmacological thromboprophylaxis. We separated bilateral primary total hip replacements performed during the same procedure (1394 in 697 patients) because of the associated increase in venous thromboembolism risk27,28. In total, data on 67,469 primary unilateral total hip replacements were available for analysis.
Study Outcome
The outcome measure was hospitalization with venous thromboembolism in the ninety days following primary total hip replacement. We obtained information from the National Registry of Patients on all patients managed with a total hip replacement who had a discharge diagnosis of deep venous thrombosis or pulmonary embolism (ICD-10 codes I26, I80.1 to I80.9, and I82.1 to I82.9). Patients who sustained a venous thromboembolism in the hospital immediately after surgery were included in the outcome analysis, while patients seen only in the emergency department were excluded. Patients with superficial venous thrombosis were not included in the analyses because they are treated as outpatients by general practitioners and would not require hospitalization in Denmark. A venous thromboembolism diagnosis is based on symptomatic events, confirmed by ultrasonography, venography, high-probability lung scanning, or computerized tomography in accordance with an existing guideline and its application29,30.
We further reported estimates of two scenarios with the highest rate of cases: (1) the incidence of venous thromboembolism including hospitalization for nonfatal symptomatic venous thromboembolism and venous thromboembolism-related deaths and (2) the incidence of venous thromboembolism including hospitalizations for nonfatal symptomatic venous thromboembolism and all deaths for the entire study population.
Statistical Analysis
Age, sex, indication for primary total hip replacement, calendar year of surgery, and level of comorbidity (Charlson comorbidity index31) were included in the analyses as possible patient-related predictors (Table I). Information on all risk factors except for comorbidity was collected from the Danish Hip Arthroplasty Register. For each patient receiving a total hip replacement, we reviewed the hospitalization history for fifteen years prior to the total hip arthroplasty and calculated a Charlson comorbidity index at the time of surgery, using all ICD-8 and ICD-10 discharge diagnoses available from the Danish National Registry of Patients. We defined three comorbidity levels: a Charlson comorbidity index of 0 (low), given to patients with no previously recorded disease categories included in the index; a score of 1 or 2 (medium); and a score of =3 (high)32 (see Appendix).
In order to examine the effect of specific comorbidities on the risk of venous thromboembolism after total hip replacement, we also categorized patients who had a total hip replacement according to the presence of given conditions at the time of surgery—for example, cardiovascular diseases (including previous myocardial infarction, congestive heart failure, and/or stroke), previous hospitalizations for deep venous thrombosis and/or pulmonary embolism, diabetes, liver and kidney diseases, cancer, and other diseases (including dementia, chronic pulmonary disease, connective tissue disease, and ulcer disease). We further examined the effect of cancer duration before surgery on venous thromboembolism risk and the distribution of cancer types in our population.
Follow-up began on the day of the primary total hip replacement and ended on the day of rehospitalization with a venous thromboembolism; on the day of discharge from the primary hospitalization for total hip replacement if venous thromboembolism occurred; or at the time of death, emigration, or ninety days after total hip replacement surgery, whichever came first.
We used the Kaplan-Meier method to calculate the ninety-day overall risk of venous thromboembolism and the median time to hospitalization with venous thromboembolism. A marginal Cox model with a common baseline hazard function (Wei-Lin-Weissfeld model) was used to examine the association between possible risk factors and time to hospitalization with venous thromboembolism, accounting for clustering within hospitals. In addition to crude relative risks, we estimated a ninety-day hazard ratio as a measure of relative risks with a 95% confidence interval for each factor, mutually adjusted for all possible risk factors. The assumption of the Cox proportional hazard model was assessed graphically with use of log-log plots and Schoenfeld residuals and was found suitable. Dose-response association was assessed by the Cochran-Armitage trend test with significance at the 0.05 level.
Separate analyses were performed to compare unilateral and bilateral one-stage total hip replacement, in which bilateral one-stage total hip replacement was included in the data set only for the purpose of this analysis.
This study was approved by the National Board of Health and the Danish Data Protection Agency.
Source of Funding
There was no external funding source.
The characteristics of the study population of 67,469 patients managed with a total hip replacement, all of whom received pharmacological thromboprophylaxis, are presented in Table I. Despite thromboprophylaxis, the overall rate of hospitalization for venous thromboembolism or discharge from the primary hospitalization for total hip replacement if venous thromboembolism occurred within ninety days after a primary total hip replacement was 1.02% (686 patients), occurring at a median of twenty-two days (25th to 75th percentile: thirteen to forty-two days) (Fig. 1). The cumulative incidence of hospitalization due to symptomatic venous thromboembolism and to nonfatal pulmonary embolism was 0.7% (499 patients) and 0.3% (205 patients), respectively, within ninety days after primary surgery. Few patients (eighteen) were discharged with a diagnosis of both venous thromboembolism and a nonfatal pulmonary embolism. Among patients hospitalized with venous thromboembolism, 15.1% were admitted to orthopaedic departments and 84.9% were admitted to other departments, mainly general internal medicine.
Of the 686 patients hospitalized with venous thromboembolism, seventy-one (10.3%) were, after initial hospitalization, treated as outpatients for the remainder of the venous thromboembolism course. Sixty-four (12.8%) of the 499 patients initially hospitalized with deep venous thrombosis were treated afterward as outpatients, whereas nine (4.4%) of 205 patients initially hospitalized with a pulmonary embolism were treated afterward as outpatients. (Two patients had both deep venous thrombosis and pulmonary embolism.)
A review of death certificates showed that 1.0% (678) of the 67,469 patients died from all causes within ninety days of surgery, but only 0.05% (thirty-eight) of the deaths were registered as related to venous thromboembolism. Thus, the overall rate of clinically manifested venous thromboembolism (nonfatal symptomatic venous thromboembolism and venous thromboembolism-related deaths) was 1.07% (686 hospitalizations and thirty-eight deaths occurring after 67,469 total hip replacements). If all causes of death were assumed to be associated with venous thromboembolism, the resultant scenario with the highest rate of clinically evident venous thromboembolism within ninety days of surgery would be 2.02% (686 hospitalizations and 678 deaths occurring after 67,469 total hip replacements) (Table II).
Patient-Related Risk Factors
Sex and age were not associated with a risk of hospitalization for venous thromboembolism (Table III). Strikingly, patients with rheumatoid arthritis had a very low overall rate of venous thromboembolism (0.54%) and an adjusted relative risk for hospitalization with venous thromboembolism of 0.47 (95% confidence interval, 0.25 to 0.90), compared with patients with primary osteoarthritis. Patients with a high Charlson comorbidity index had an elevated overall rate for venous thromboembolism (1.38%) and an increased relative risk for venous thromboembolism compared with patients with a low Charlson comorbidity index (adjusted relative risk = 1.45; 95% confidence interval, 1.02 to 2.05).
The overall rate of hospitalization for venous thromboembolism among patients managed with a total hip replacement who were previously hospitalized with cardiovascular disease was 1.35%. Their relative risk of venous thromboembolism was elevated relative to patients who had not been hospitalized with these diseases (adjusted relative risk = 1.40; 95% confidence interval, 1.15 to 1.70). Patients previously hospitalized with other diseases, including dementia, chronic pulmonary disease, connective tissue disease, and ulcer disease, had an overall rate for venous thromboembolism of 1.29%, resulting in an adjusted relative risk of 1.45 (95% confidence interval, 1.21 to 1.72) compared with patients with no previous hospitalization with these diseases. Finally, patients previously hospitalized with venous thromboembolism had an overall risk of hospitalization for a new venous thromboembolism of 7.52% and an adjusted relative risk of 8.09 (95% confidence interval, 6.07 to 10.77) compared with patients with no previous venous thromboembolism-related hospitalization.
A previous hospitalization with any type of cancer did not appear to influence the risk of venous thromboembolism after total hip replacement (adjusted relative risk = 0.93; 95% confidence interval, 0.68 to 1.28) (Table III). The lowest risk was found among patients diagnosed with cancer within one year before total hip replacement (adjusted relative risk = 0.39; 95% confidence interval, 0.10 to 1.57), while patients hospitalized with cancer more than a year before total hip replacement had an adjusted relative risk for venous thromboembolism of 1.09 (95% confidence interval, 0.83 to 1.42). A total of 4785 patients having a total hip replacement (7.1%) had been hospitalized with a cancer diagnosis before surgery, most frequently stomach, colon, or rectal cancer (22%), genital cancer (22%), or breast cancer (14%).
There was an indication of an increasing overall risk of hospitalization for venous thromboembolism during the study period (p = 0.0004 for test for trend). Thus, adjusted relative risks were 0.98 (95% confidence interval, 0.75 to 1.28), 1.08 (95% confidence interval, 0.79 to 1.48), and 1.25 (95% confidence interval, 0.94 to 1.67) when patients who underwent total hip replacement in 1998, 1999, or 2000; in 2001, 2002, or 2003; and in 2004, 2005, or 2006 were compared with patients who had the procedure in 1995, 1996, or 1997 (Table III). We found no difference in venous thromboembolism risk among patients receiving the three most frequently used thromboprophylactic drugs, including dalteparin (Fragmin), enoxaparin (Klexane), tinzaparin (Innohep), and a mixed group of other drugs. However, there was a significantly lower risk among patients treated with fondaparinux (Arixtra) (Table III).
We compared the risk for venous thromboembolism between bilateral one-stage total hip replacement (six of 697 procedures) and unilateral total hip replacement (686 of 67,469 procedures), resulting in an adjusted relative risk for venous thromboembolism of 1.14 (95% confidence interval, 0.13 to 8.21).
In this observational population-based study, the cumulative incidence of a venous thromboembolism within ninety days of surgery among 67,469 patients with a total hip replacement receiving pharmacological thromboprophylaxis was approximately 1%. The overall incidence of venous thromboembolism did not decline during the ten-year study period. Several risk factors for venous thromboembolism were identified.
Study Strengths and Limitations
The main strengths of our study were its population-based prospective design with complete follow-up and a large sample size. In addition, a previous study33 has described high validity of the data in the Danish Hip Arthroplasty Register, with 94% completeness of the procedures registered and a positive predictive value of between 84% and 100% for hip diagnoses for primary total hip replacement. As well, our study was based on data colected independently of its objective. Although data were missing for about 6% of the patients who had a total hip replacement, this is likely to be independent of later registration with venous thromboembolism. Completeness and quality of other risk factors collected in Danish health-care registries is as high as 100% because of comprehensive quality control every three months.
The validity of our findings also depends on accurate coding of venous thromboembolism in the National Registry of Patients. The predictive value of a discharge diagnosis of venous thromboembolism in hospital discharge registries previously has been reported to be 80% to 90%34-37. While patients who died within ninety days of total hip replacement were excluded from the analyses, we addressed this issue in sensitivity analyses. A potential study limitation is the lack of access to autopsy results, which could produce more conservative estimates of overall venous thromboembolism risk. Because the lack of autopsy tests is most likely nondifferential, this should not have affected the relative risk estimates.
Further, we were not able to differentiate patients with a history of hospitalization for cancer from patients with an active neoplastic process concurrent with total hip replacement. Since the patient with active cancer may be of more interest in relation to the risk for venous thromboembolism, we addressed this issue in part, presenting estimates of the risk for venous thromboembolism separately for patients diagnosed with cancer within one year before total hip replacement and patients hospitalized with cancer more than a year before total hip replacement, using the date of diagnosis as a proxy measure of current cancer activity. Since we did not have information on the type of thromboprophylactic regimen used in relation to specific drugs, our risk estimates, in terms of the type of pharmacological thromboprophylaxis, may be biased. As well, although we performed multivariate analyses, unmeasured (for example, lack of data on physical activity) and residual confounding remains a general threat to all observational studies.
Overall Incidence of Venous Thromboembolism
The overall incidence of venous thromboembolism in our study accords with the findings of one randomized trial9, but is lower than that found in most previous observational and randomized studies7,11,14,15,18,19,38,39. The discrepancy may stem from differences in study populations, study design, type of data sources, proportion of patients receiving pharmacological thromboprophylaxis, and type and duration of treatment. However, the main reason for the discrepancy is probably the definition of outcomes and the methods of outcome ascertainment—for example, whether there is routine screening of the study population for venous thromboembolism. The separate incidences of venous thromboembolism and pulmonary embolism of 0.7% and 0.3%, respectively, in our study were lower than the incidence of venous thromboembolism ranging from 1.1% to 3.4% in several studies9,14,19,38,39 and lower than the incidence of pulmonary embolism of 1.1% reported in three studies38-40. Nevertheless, the overall risk of venous thromboembolism in orthopaedic patients is high, given the previous research showing that >25% of patients die within one year after the onset of venous thromboembolism41.
Patient-Related Risk Factors
The finding that the age of the patient receiving a total hip replacement did not affect the risk of venous thromboembolism agrees with the results of some14,17, but not all, earlier studies18. Age was found to be associated with an increase in venous thromboembolism among patients with a low comorbidity index who had an appendectomy but not among patients with a high comorbidity index who had vascular surgery, which is similar to previous findings among patients who had a total hip replacement42. Thus, it appears that, in relation to venous thromboembolism, comorbidity is a stronger confounder than age.
An interesting study finding is the reduced risk of venous thromboembolism among patients with rheumatoid arthritis who had a total hip replacement compared with those with osteoarthritis. In general, patients with rheumatoid arthritis have an increased burden of arterial cardiovascular events43, probably because of acceleration and triggering of atherogenesis. It appears that the mechanisms underlying venous thromboembolism and arterial cardiovascular events differ in patients with rheumatoid arthritis probably because of different risk factors or pathophysiological entities, although recent research has indicated an association between venous thromboembolism and atherosclerotic disease in general44. We cannot entirely exclude the possibility of confounding by indication or unmeasured confounding, however. For example, patients with rheumatoid arthritis are more likely to receive several potent medications (such as methotrexate, sulfasalazine, and steroids) because of their underlying disease than are patients with osteoarthritis. The impact of these drugs on the venous thromboembolism risk is unclear.
The finding of an increased risk for venous thromboembolism in patients with a high comorbidity index who are managed with total hip replacement accords with the findings of Lyman et al.45 and Mantilla et al.17. We further confirm that a history of thromboembolism increases the risk for venous thromboembolism after surgery14,17,18. The higher risk may reflect a higher likelihood for a clot to form because of previous endothelial damage or the presence of one or more underlying genetic or acquired defects of the coagulation system46. Nevertheless, patients receiving a total hip replacement are already at high baseline risk of sustaining a venous thromboembolism because of prolonged immobility, with direct venous trauma and venous stasis.
We found no difference in venous thromboembolism risk in patients with cancer compared with those without a cancer history, which diverges from current thinking47. Only a few studies have adjusted for other important risk factors, such as age and comorbidity, which interact closely with cancer48. No study that we are aware of has so far addressed the effect of malignancy on the risk of venous thrombosis following total hip replacement. Nonetheless the issue of patient selection may blur our estimated risk; cancer patients may have a lower likelihood of undergoing total hip replacement than other patients.
Numerous studies recommending new and more effective pharmacological and nonpharmacological prevention of postoperative venous thromboembolism have been published in the last decade. However, we were unable to identify a decline in the overall risk of postoperative venous thromboembolism since 1995; on the contrary, the risk appeared to be increasing, in line with the findings of some49, but not all, studies38. It is worth noting, however, that revised guidelines50 for extended thromboprophylaxis in patients undergoing total hip replacement have not yet been fully implemented in clinical practice in Denmark. At the same time, the duration of hospitalization for total hip replacement continuously declined during the study period because of the implementation of so-called fast-track surgery in Denmark51.
In conclusion, we found that one of every 100 patients managed with a total hip replacement who received pharmacological thromboprophylaxis manifested clinically evident venous thromboembolism within ninety days of surgery. We identified several subgroups of patients at an elevated risk for hospitalization with postoperative venous thromboembolism, including patients with a high Charlson comorbidity index and those previously hospitalized with cardiovascular or cerebrovascular diseases or venous thromboembolism. Patients with rheumatoid arthritis had a reduced risk for hospitalization with postoperative venous thromboembolism. The incidence of venous thromboembolism following total hip replacement did not decline in the ten-year study period. This knowledge can be used in the future to better stratify patients having a total hip replacement according to the risk for venous thromboembolism and to target specific thromboprophylaxis modalities to these patients.