Figure 1. Adjusted hazard ratios (HRs) for acute myocardial infarction (AMI). THR indicates total hip replacement; TKR, total knee replacement.
Figure 2. Cumulative incidence rates of acute myocardial infarction (AMI). A, Patients undergoing total hip replacement (THR); B, patients undergoing total knee replacement (TKR).
Arief Lalmohamed, Peter Vestergaard, Corinne Klop, Erik Lerkevang Grove, Anthonius de Boer, Hubertus G. M. Leufkens, Tjeerd P. van Staa, Frank de Vries. Timing of Acute Myocardial Infarction in Patients Undergoing Total Hip or Knee ReplacementA Nationwide Cohort Study. Arch Intern Med. 2012;172(16):1229–1235. doi:10.1001/archinternmed.2012.2713
Author Affiliations: Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht Institute for Pharmaceutical Sciences, Utrecht University, Utrecht, the Netherlands (Mr Lalmohamed, Ms Klop, and Drs de Boer, Leufkens, van Staa, and de Vries); Departments of Endocrinology and Internal Medicine (Dr Vestergaard) and Cardiology (Dr Grove), Aarhus University Hospital, Aarhus, Denmark; MRC Lifecourse Epidemiology Unit, Southampton General Hospital, Southampton, England (Drs van Staa and de Vries); and Department of Clinical Pharmacy and Toxicology, Maastricht University Medical Centre, Maastricht, the Netherlands (Dr de Vries). Dr Vestergaard is now with the Medical Faculty, Aalborg University, Aalborg, Denmark.
Background Limited evidence suggests that the risk of acute myocardial infarction (AMI) may be increased shortly after total hip replacement (THR) and total knee replacement (TKR) surgery. However, risk of AMI in these patients has not been compared against matched controls who have not undergone surgery. The objective of this study was to evaluate the timing of AMI in patients undergoing THR or TKR surgery compared with matched controls.
Methods Retrospective, nationwide cohort study within the Danish national registries. All patients who underwent a primary THR or TKR (n = 95 227) surgery from January 1, 1998, through December 31, 2007, were selected and matched to 3 controls (no THR or TKR) by age, sex, and geographic region. All study participants were followed up for AMI, and disease- and medication history–adjusted hazard ratios (HRs) were calculated.
Results During the first 2 postoperative weeks, the risk of AMI was substantially increased in THR patients compared with controls (adjusted HR, 25.5; 95% CI, 17.1-37.9). The risk remained elevated for 2 to 6 weeks after surgery (adjusted HR, 5.05; 95% CI, 3.58-7.13) and then decreased to baseline levels. For TKR patients, AMI risk was also increased during the first 2 weeks (adjusted HR, 30.9; 95% CI, 11.1-85.5) but did not differ from controls after the first 2 weeks. The absolute 6-week risk of AMI was 0.51% in THR patients and 0.21% in TKR patients.
Conclusions Risk of AMI is substantially increased in the first 2 weeks after THR (25-fold) and TKR (31-fold) surgery compared with controls. Risk assessment of AMI should be considered during the first 6 weeks after THR surgery and during the first 2 weeks after TKR surgery.
Total hip replacement (THR) and total knee replacement (TKR) are highly effective in patients with moderate to severe osteoarthritis.1 These surgical procedures are frequently performed, yielding an estimated annual number of 1.8 million procedures worldwide.2,3 Among patients undergoing THR or TKR surgery, acute myocardial infarction (AMI) has been identified as an important perioperative complication.4,5 In the general population, AMI is a major cause of morbidity and mortality worldwide,6 and each year more than 7 million patients are estimated to sustain an AMI.6 In THR or TKR patients, the risk of AMI may be decreased or increased shortly after surgery. On one hand, the surgery itself may result in ischemic complications caused by marrow embolization.7,8 On the other hand, antithrombotic agents are commonly used in these patients during hospitalization and have the potential to decrease the risk of AMI.9 Epidemiologic studies4,10- 15 have reported 90-day AMI rates of up to 1.8%, of which most occurred within the first week.
Timing of AMI after THR or TKR surgery has become of increasing interest.14 Although early hospital discharge has been promoted in these patients, perioperative complications, including AMI, may argue against this practice.16 Because no previous studies included a large control cohort for reference, it is thus difficult to interpret the magnitude of increased AMI risk after THR or TKR surgery compared with the general population. Differences in baseline characteristics among the studies further add to this difficulty. More important, previous studies have only focused on short-term AMI risk (ie, <90 days) and did not investigate long-term risk for AMI.
Furthermore, data are limited on individual risk factors for AMI after THR or TKR surgery. This drawback is of particular importance given the number of comorbidities often present in these elderly patients. Previous studies4,10- 15 were limited by several design issues, such as small sample sizes and lack of matched control cohorts who did not undergo THR or TKR surgery. Moreover, none of these studies provided analyses adjusted for medication. For example, use of pain relievers (in particular, nonsteroidal anti-inflammatory drugs [NSAIDs]) is common among THR and TKR patients and might increase the risk of AMI.17,18 The objectives of this study were to evaluate the timing of AMI after THR and TKR surgery, to evaluate potential effect modifiers of this relationship, and to identify determinants of AMI in THR and TKR patients.
Using Danish national registries, we conducted a nationwide retrospective cohort study. The total population from which the study participants were drawn was 5.5 million. Detailed information was available for all Danish residents, including data on second-line visits (hospitals, outpatient clinics, and emergency departments; from 1977 onward), drugs sold at retail pharmacies (from 1996 onward), citizen status (vital status, date of death, residence, migration, and socioeconomic status; from 1968 onward), and causes of death (1 underlying cause and up to 3 additional immediate causes; from 1970 onward). In Denmark, all residents have free access to health services, including hospital services and visits to general practitioners (tax funded). Previous reports demonstrated high quality, completeness, and validity rates, and these registries have been used in numerous recent epidemiologic studies.19
All patients 18 years or older who underwent a primary THR or primary TKR from January 1, 1998, through December 31, 2007, were included in the study. Both THR and TKR were identified using hospital discharge records and were classified by the International Classification of Diseases, 10th revision(ICD-10)20 (ICD-10 code NFB for THR and ICD-10 code NGB for TKR). Each THR and TKR patient was matched with 3 controls of the same age and sex without a history of THR and TKR. The index date was defined as the date of primary THR and TKR hospital admission for THR and TKR patients and similarly for matched controls. We excluded individuals with a prior AMI within 6 weeks before or on the index date.
Danish guidelines recommend thromboprophylaxis (mostly low-molecular-weight heparin [LMWH]; started 12 hours before surgery or 12-24 hours after surgery) for all THR and TKR patients while in the hospital, which can be extended up to 35 days.21 Previous Danish data revealed that 99.1% of THR and TKR patients had indeed received thromboprophylactic agents (of which 93% included LMWHs).22
All patients were followed up from the index date until death, migration, THR or TKR revision, or the end of the study period (December 31, 2007) or AMI, whichever came first. Acute myocardial infarction was assessed using the National Hospital Discharge Registry and the Danish Causes of Death Registry (both classified using ICD-10 code I21). Acute myocardial infarction was divided into fatal and nonfatal events based on death certificates.
We reviewed the literature to define potential (general) risk factors and confounders for this study.23,24 These factors included age, sex, socioeconomic status, indication for surgery, a history of AMI (stratified by time between most recent AMI and THR or TKR surgery), history of other ischemic heart disease, heart failure, and cerebrovascular disease. Furthermore, a drug dispensing for β-blockers, renin-angiotensin-aldosterone system inhibitors, thiazide diuretics, calcium channel blockers, organic nitrates, statins, nonselective NSAIDs (including high-dose aspirin), cyclo-oxygenase 2 selective inhibitors, antiplatelet drugs, vitamin K antagonists, estrogen-containing drugs, antidiabetic drugs, and inhaled β2-agonists within 6 months were considered as potential confounders for AMI.
Using the PHREG procedure from SAS statistical software, version 9.2 (SAS Institute, Inc), we calculated hazard ratios (HRs) for the risk of AMI with THR and TKR and compared them with age- and sex-matched controls (stratified on matched pairs). Total follow-up time was divided into 6-week periods and the first 6 weeks into 1-week periods. Information on potential confounders and risk factors was collected during follow-up; before the start of each period, we evaluated the presence of these covariates. Potential confounders were included in the final model if they independently changed the β-coefficient for THR or TKR by at least 5%.
To assess the timing of AMI after THR and TKR surgery, we included period interaction terms (period × surgery) in the model for the following periods: less than 2 weeks, 2 to 6 weeks, 6 to 12 weeks, 3 to 6 months, 6 to 12 months, and 1 year or more after surgery. For each period, AMI risk was plotted against the median time since THR or TKR surgery and visualized using smoothing spline regression,25- 28 which has been advocated as an alternative to categorical analysis.29 In addition, we used Kaplan-Meier plots to present the cumulative incidence rates of AMI over time (divided into fatal and nonfatal events).
To compare AMI risk after THR or TKR surgery with other elective operations, we performed a sensitivity analysis. Within THR matched controls, we selected patients who underwent hernia surgery. For these controls, the index date was reset at time of elective surgery hospital admission. The THR patients whose matched controls did not undergo these elective operations were excluded, and the analyses were further adjusted for calendar year, sex, and age at surgery.
For potential effect modifiers and determinants, we evaluated 2 periods by restricting follow-up to less than 6 weeks or 6 to 52 weeks after surgery. Potential effect modifiers were screened by entering an interaction term (risk factor × surgery) into the model. To identify determinants of AMI within THR and TKR patients only, we excluded controls and used stepwise backward elimination to determine the final regression model after entering all previously mentioned risk factors (P < .05) into the model. This study was approved by the National Board of Health and the Danish Data Protection Agency.
After exclusion of 437 patients with an AMI in the 6 weeks before or on the index date, 66 524 THR patients, 28 703 TKR patients, and 286 165 matched controls were enrolled in the study (Table 1). Because of matching, patients had a similar distribution of age (THR: mean age, 71.9 years; TKR: mean age, 67.2 years) and sex (THR: 36.9% male; TKR: 37.6% male) compared with matched controls. The THR and TKR patients were more likely to have used NSAIDs compared with controls and had slightly more often been diagnosed as having ischemic heart disease before surgery.
Figure 1 shows that the risk of AMI was substantially increased during the first 2 weeks after THR or TKR surgery compared with controls. Adjusted HRs were 25.5 (95% CI, 17.1-37.9) for THR and 30.9 (95% CI, 11.1-85.5) for TKR. Compared with patients who underwent hernia surgery, the 2-week AMI risk remained significantly elevated (adjusted HR, 21.9; 95% CI, 2.94-163.2). In TKR patients, the risk reached baseline levels after the first 2 weeks (2-6 weeks: adjusted HR, 0.81; 95% CI, 0.37-1.77), whereas in THR patients, the risk remained elevated during the first 6 weeks after surgery (2-6 weeks: adjusted HR, 5.05; 95% CI, 3.58-7.13). Kaplan-Meier plots revealed the same timing patterns (Figure 2). Absolute 6-week rates of AMI were 0.51% for THR patients and 0.21% for TKR patients.
For both THR and TKR, we found a strong effect modification by age (Table 2). During the first 6 weeks, the effect of THR on AMI risk was highest in the oldest patients (≥80 years old; adjusted HR, 25.3; 95% CI, 17.7-36.2), whereas we could not detect a significantly increased risk in patients younger than 60 years (adjusted HR, 2.41; 95% CI, 0.68-8.57). We found a similar, albeit less substantial, age trend with TKR surgery. No other significant effect modifiers for the relationship between THR or TKR and AMI during the first 6 postoperative weeks were identified.
In the THR patients, the 6-week risk of AMI was higher among older patients; men; patients with a previous AMI, heart failure, or cerebrovascular disease; and users of NSAIDs, β-blockers, potassium-sparing diuretics, organic nitrates, and antiplatelet drugs during follow-up compared with THR patients without these characteristics (Table 3). The elevated risk caused by a previous AMI before THR or TKR surgery diminished with an increasing time since most recent AMI before surgery (Table 3).
This study demonstrated an increased risk of AMI during the first 2 weeks after THR (25-fold) and TKR (31-fold) surgery compared with matched controls. The risk of AMI sharply decreased after this period, although it remained significantly elevated in the first 6 weeks for THR patients. The association was strongest in patients 80 years or older, whereas we could not detect a significantly increased risk in patients younger than 60 years. Furthermore, a previous AMI in the 6 months before surgery increased the risk of new AMI during the first 6 weeks after THR and TKR (4-fold increase) surgery but did not modify the relationship between THR or TKR and AMI.
To our knowledge, this is the first study comparing AMI risk after THR or TKR surgery with the risk of matched controls not undergoing surgery. Previous studies were limited to reports on (primarily perioperative) incidence rates only and showed somewhat conflicting results. For example, Khatod et al11 demonstrated a 0.1% incidence rate of AMI within 90 days after TKR surgery, whereas Gandhi et al14 found a 1.8% incidence rate in the first 18 days after THR or TKR surgery. This discrepancy may partially be explained by differences in diagnosing AMI because the latter study used serum troponin levels in addition to electrocardiogram changes for diagnosis. Most other studies4,12,13,15 found an AMI incidence rate of 0.3% to 0.8%, which is well in line with our findings. Because most of these studies included perioperative events only (typically <20 days), our incidence rates tended to be more toward 0.8% rather than the lower end. Alternatively, the discrepancy may be explained by differences in baseline characteristics among the studies, including comorbid cardiovascular disease and characteristics of the orthopedic center performing the surgical procedure. An American study30 thus showed that high-volume hospitals had a lower 30-day mortality rate after major orthopedic surgery, although no adjustments were made for comorbidities or surgical complexity.
Evidence on timing of AMI after THR and TKR surgery is scarce. Previous studies have only found an elevated risk during the first 4 to 5 postoperative days. Gandhi et al14 found that within 5 days after THR or TKR surgery, 91% of all in-hospital AMI events had occurred. Similarly, Parvizi et al13 found that perioperative AMIs were most likely to occur within 4 days after THR or TKR surgery. Our findings confirm this increased risk of AMI and suggest that the risk is actually increased for an even longer period (THR: first 6 weeks; TKR: first 2 weeks).
The biological mechanism explaining the increased risk of AMI may be related to marrow embolization because surgical invasion of the medullary canal of the femur potentially causes marrow embolization and cardiac stress.14 This embolization process occurs primarily with THR and to a lesser extent with TKR.7,8 This fact may explain the differences in AMI risk between THR and TKR observed in our study. Among THR patients, the increase in AMI risk lasted for a longer period compared with TKR patients. Furthermore, hemodynamic stressors associated with the surgery (eg, effects of anesthesia on the cardiovascular system, blood loss, fluid shifts, arrhythmias, and hypoxia) can further contribute to the observed increased risk of AMI after THR and TKR surgery.
It is unlikely that the use of inpatient antithrombotic agents will explain the observed elevated risk of AMI after THR and TKR surgery. Most Danish THR and TKR patients are treated with LMWHs,22 which have been shown to lower the risk of death and myocardial infarction during the first 6 days of therapy in patients with unstable coronary artery disease.9 This finding would imply that we may have underestimated the risk of AMI and that the actual association between THR or TKR and risk of AMI would be even stronger. There is conflicting evidence about the association between dabigatran etexilate and an increased risk of AMI.31 However, dabigatran was not available during the entire study period and should therefore not have influenced our results. As a further note, patients taking (outpatient) antithrombotic agents may represent a higher-risk population (eg, use of low-dose aspirin to prevent secondary events). This may have cancelled our effect modification and is most likely the reason why antiplatelet drugs were identified as a significant determinant of AMI during the first 6 weeks after THR and TKR surgery.
Our study implies that a recent AMI (within 1 year) should be a contraindication for those undergoing elective THR surgery. Previous literature confirmed AMI as a risk factor for a new AMI in these patients.14 However, no other study has evaluated the time since most recent AMI, but this is important when planning the performance of THR. We were able to show a sharp decrease in risk of a new AMI when the previous AMI had occurred more than 1 year before surgery. However, even beyond this period, the risk remained elevated compared with those without a previous AMI. These findings are indirectly supported by a Swedish retrospective cohort study32 in patients with ST-elevation myocardial infarction. The authors of that study reported that the risk of reinfarction was highest within the first year of AMI.
Strengths of this study include the nationwide population-based design, the large sample size, information on matched controls, and completeness of follow-up. Unlike most other studies, we had access to outpatient prescription data (such as NSAIDs) and information from outpatient clinics. Because we had highly valid data on mortality, we were able to identify out-of-hospital fatal AMI events. The major drawback is the lack of information on other risk factors for AMI, such as smoking, blood pressure, biochemical variables, and body mass index. A higher body mass index is associated with an increased risk of coronary artery disease33 and osteoarthritis, the main indication for THR and TKR. However, in a previous study34 on patients undergoing THR, body mass index at the time of surgery was not associated with short- or long-term mortality. Furthermore, we did not have information on inpatient anticoagulant use. Because warfarin and LMWHs have been shown to reduce AMI incidence, this could have distorted our study findings.9,35As explained, this would mean an underestimation of our observed increased AMI risk. We cannot exclude the possibility that hospitalized patients may have been more likely to be diagnosed as having an AMI. However, we did not look at silent myocardial infarctions (which are more likely to be recorded as silent ischemic events rather than AMIs). Moreover, we also found an increased risk of fatal AMIs, for which the detection rate should be equal. Finally, we did not have information about general anesthesia, which may well be the cause of the increased risk of AMI after THR and TKR surgery. However, a previous study36 that evaluated the influence of general anesthesia in surgical patients vs those who received regional anesthesia showed a trend toward only a 1.4-fold increased risk of AMI. This is well below the excess risk we observed in our study, suggesting that the increased risk in THR/TKR patients might not be fully explained by general anesthesia only. Furthermore, our sensitivity analysis demonstrated that the increased risk of AMI after THR surgery remained elevated when compared with other elective operations.
To our knowledge, this is the first study that found that THR (25-fold) and TKR patients (31-fold) are at increased risk of AMI during the first 2 weeks after surgery. The elevated risk was sustained for 6 weeks after THR and for 2 weeks after TKR. The effect of surgery on AMI risk was strongest in patients 80 years or older. The relationship was not more pronounced in those with well-known risk factors of AMI (such as heart failure, cerebrovascular disease, and previous AMI), although they increased the risk of AMI within THR and TKR patients. Finally, our data suggest that elective THR surgery should be contraindicated in patients with a previous AMI in the last 12 months before surgery.
Correspondence: Frank de Vries, PharmD, PhD, Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht Institute for Pharmaceutical Sciences, Utrecht University, PO Box 80082, 3508 TB Utrecht, the Netherlands (firstname.lastname@example.org).
Accepted for Publication: May 6, 2012.
Published Online: July 23, 2012. doi:10.1001/archinternmed.2012.2713
Author Contributions:Study concept and design: Lalmohamed, Vestergaard, de Boer, Leufkens, and de Vries. Acquisition of data: Lalmohamed, Vestergaard, and de Vries. Analysis and interpretation of data: Lalmohamed, Klop, Grove, de Boer, Leufkens, van Staa, and de Vries. Drafting of the manuscript: Lalmohamed and de Vries. Critical revision of the manuscript for important intellectual content: Lalmohamed, Vestergaard, Klop, Grove, de Boer, Leufkens, van Staa, and de Vries. Statistical analysis: Lalmohamed, Vestergaard, and de Boer. Obtained funding: Lalmohamed and Vestergaard. Administrative, technical, and material support: Vestergaard, Leufkens, and de Vries. Study supervision: Vestergaard, Grove, de Boer, Leufkens, van Staa, and de Vries.
Financial Disclosure: The Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht Institute for Pharmaceutical Sciences, which employs Mr Lalmohamed, Ms Klop, and Drs de Boer, Leufkens, van Staa, and de Vries, has received unrestricted funding for pharmacoepidemiologic research from GlaxoSmithKline, the private-public–funded Top Institute Pharma (www.tipharma.nl; includes co-funding from universities, government, and industry), the Dutch Medicines Evaluation Board, and the Dutch Ministry of Health.
Funding/Support: This study was supported by grant 017.007.010 from the Netherlands Organization for Scientific Research (Nederlandse Organisatie voor Wetenschappelijk Onderzoek, The Hague, the Netherlands).