Total knee arthroplasty is among the most common orthopaedic procedures, with over a half million procedures being performed in the United States annually1. In spite of this commonality, the best method for providing anesthesia and analgesia during arthroplasty procedures has not been established.
Compared with general anesthesia, neuraxial anesthesia in orthopaedic procedures has been reported to decrease the rates of deep-vein thrombosis, pulmonary embolism, and intraoperative bleeding; the need for transfusion; the length of hospital stay; the risk of surgical site infection; and the total operative cost2-8. Spinal anesthesia is a type of neuraxial anesthesia. Although it is less extensively studied in orthopaedic patients, some studies have demonstrated no significant difference between spinal and general anesthesia in terms of complication rates9,10. Also, spinal anesthesia carries a unique set of risks, including an increased potential for paresthesias and neurologic injury11. Furthermore, comparisons between these studies is limited by small patient numbers5. To our knowledge, no multicenter, prospective comparison of complications between spinal and general anesthesia during total knee arthroplasty has been conducted.
The purpose of the present study was to identify differences in thirty-day perioperative morbidity and mortality between patients managed with spinal and general anesthesia at the time of total knee arthroplasty. We searched a database that is prospectively maintained by the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) to identify all total knee arthroplasty procedures that were performed between 2005 and 2010 and hypothesized that significant differences would exist between the two anesthetic methods.
Materials and Methods
All patients who underwent a primary total knee arthroplasty between 2005 and 2010 were identified from the ACS NSQIP Participant Use Data Files. This database has been widely used to investigate short-term surgical outcomes12-20. The details of the NSQIP data-collection methods are beyond the scope of this article but can be found elsewhere21. In brief, Surgical Clinical Reviewers at each participating institution prospectively collect preoperative and postoperative morbidity and mortality data on each patient. Data collection continues on an outpatient basis, and internal auditing ensures high-quality data with an overall disagreement rate of <1.8%. The database contains patient data from 258 hospitals around the United States and thus is broadly generalizable. A search for the Current Procedural Terminology (CPT) code 27447 (arthroplasty, knee, condyle and plateau; medial and lateral compartments with or without patellar resurfacing) identified 15,489 cases. Finally, only patients undergoing either general anesthesia (8022 patients) or spinal anesthesia (6030 patients) (total, 14,052 patients) were included in our analysis. Patients who received another anesthetic type, such as regional anesthesia, were excluded.
Complications after total knee arthroplasty were defined as surgical site infection (subdivided into superficial, deep, and organ space), wound dehiscence, pulmonary embolism, deep-vein thrombosis, respiratory complications (pneumonia and unplanned intubation), renal failure/insufficiency (an increase in creatinine of >2 mg/dL above baseline), acute myocardial infarction, cardiac arrest requiring resuscitation, stroke, coma (lasting for more than twenty-four hours), peripheral nerve injury, urinary tract infection, bleeding requiring transfusion, implant failure, sepsis, septic shock, return to the operating room within thirty days after the procedure, and mortality within thirty days after the procedure.
The NSQIP database has strict definitions for each complication; the complete list is included in the User Guide21. Some important definitions follow. Superficial surgical site infection is defined as an infection that involves only skin or subcutaneous tissues and is associated with purulent drainage; organisms isolated from an aseptic culture; and/or pain, tenderness, localized swelling, redness, or heat at the site of an incision that was performed by the surgeon (with the exclusion of stitch abscesses). A deep infection involves deep soft tissues such as fascia or muscle. An organ space infection is an infection outside of the path of the incision, such as an abscess, that requires operative intervention. Bleeding transfusions are defined as the administration of any packed or whole red blood-cell product to a patient up to seventy-two hours after leaving the operating room. Graft/implant failure is defined as the mechanical failure of any extracardiac graft or prosthesis. Deep-vein thrombosis is defined as a venous blood clot diagnosed with a duplex ultrasonography, venogram, or computed tomography (CT) scan and treated with anticoagulation or vena cava filter. Peripheral nerve injuries are defined as only motor deficits that are attributable to injury to the cervical plexus, brachial plexus, ulnar plexus, sciatic, or femoral nerves.
Patient characteristics available in the NSQIP database were divided into demographic characteristics, preoperative comorbidities, preoperative laboratory values, and operative variables. As listed in Table I, demographic characteristics included age, sex, and race. Comorbidities included body mass index (BMI), recent weight loss, diabetes mellitus, smoking, alcohol use (more than two drinks per day), chronic obstructive pulmonary disease, coronary artery disease (myocardial infarction and congestive heart failure), previous transient ischemic attack, dialysis, steroid use, bleeding disorder (any condition requiring hospitalization because of deficiency of blood clotting elements), preoperative blood transfusion, open wound, chemotherapy (less than thirty days ago), radiation therapy (less than ninety days ago), recent operation (within thirty days), and preoperative sepsis. Laboratory values included the white blood-cell count, hematocrit, platelet count, creatinine, albumin, and international normalized ratio. Finally, operative variables included wound class, American Society of Anesthesiologists (ASA) class, intraoperative blood transfusions, the duration of the surgical procedure, and resident involvement (scrubbed on the case).
Unadjusted complication rates were calculated for the general and spinal anesthesia groups. A standard t test and chi-square analysis were used to identify significant differences in unadjusted individual and summated complication rates (Table II). Next, all preoperative patient demographic characteristics, preoperative comorbidities, laboratory values, and operative variables were compared between the spinal and general anesthesia groups with use of chi-square tests for categorical variables and Wilcoxon rank-sum tests for continuous variables (Table I). The multivariate model was created with use of both historic and cohort-identified risk factors. The variables in Table II were compared between patients with and without thirty-day complications. Serum albumin had <80% chart completion and was excluded from the multivariate analysis. All other variables were included. A p value of <0.1 was set to identify variables for inclusion in the multivariate model. Variables meeting this criterion included age, race, sex, BMI, diabetes mellitus, alcohol use, peripheral vascular disease, a bleeding disorder, chemotherapy, recent operation, hematocrit, creatinine, and ASA class. Additional risk factors, including smoking, coronary artery disease (recent myocardial infarction or congestive heart failure), open wound, white blood-cell count, and wound class, were included on the basis of historic recognition in the literature. NSQIP characteristics that were deemed less relevant to orthopaedic surgery, such as esophageal varices, ascites, and pregnancy, were excluded at the authors’ discretion. Before inclusion in the final model, Pearson correlation coefficients verified that no colinearity existed between variables. A multivariate logistic regression analysis requires complete patient data, and any patient with a missing data point is necessarily excluded. After the exclusion of patients with incomplete data, the final cohort size for the logistic regression analysis was 11,988 patients (including 5370 patients managed with spinal anesthesia and 6618 patients managed with general anesthesia).
Propensity scores were introduced as a method to control for selection bias between the spinal and general anesthesia groups. Three methods of propensity score analysis have been described: matching, stratification, and logistic regression22. Our study utilized two separate models: quintile stratification and propensity score-adjusted logistic regression. The propensity score was defined as the conditional probability of receiving general anesthesia based on observed covariants. The propensity score distribution among patient cohorts is shown in a figure in the Appendix.
A multivariate model involving logistic regression analysis was created from all previously identified dependent variables and the propensity score. After adjustment, the p values for preoperative comorbidities were calculated (Table I). The model successfully reduced selection bias by eliminating significant differences in group preoperative variables (p > 0.05 for all). Adjusted odds ratios with 95% confidence intervals were calculated (Table III). Finally, a separate propensity score stratification model was utilized (see Appendix). All patients from each cohort were divided into five equal strata on the basis of their propensity scores. This means that 20% of patients with the lowest propensity score were placed in the first stratum, whereas the 20% of patients with the highest scores were placed in the highest (fifth) stratum. After equal division by scores, complication rates within each stratum were calculated. Propensity matching into quintiles alone has been shown to reduce bias by 90%22. SAS for Windows (version 9.3; SAS Institute, Cary, North Carolina) was used to perform the statistical analysis. The level of significance was set at p < 0.05.
Source of Funding
No external funding was used in this study.
From the NSQIP database, 14,052 patients who underwent primary total knee arthroplasty were identified. Of these, 8022 (57.1%) were managed with general anesthesia and 6030 (42.9%) were managed with spinal anesthesia. As seen in Table II, unadjusted direct comparison of the patient cohorts revealed that the rate of complications (10.72% versus 12.34%; p = 0.0032), the rate of superficial wound infections (0.68% versus 0.92%; p = 0.0003), the rate of blood transfusions (5.02% versus 6.07%; p = 0.0086), the length of the surgical procedure (ninety-six versus 100 minutes; p < 0.0001), and the duration of the hospital stay (3.45 versus 3.77 days; p < 0.0001) were all lower in the spinal anesthesia group as compared with the general anesthesia group.
Significant differences were observed between the spinal and general anesthesia groups in terms of preoperative characteristics (Table I). To account for potential bias introduced by these differences, we performed two separate statistical corrections with use of propensity scores: stratification and logistic regression. Stratification allowed matching of patient cohorts with similar covariants. Patients in quintile 5 (the 20% of each cohort with the highest conditional probability of complication based on the highest number of covariants and highest ASA class) maintained a significantly lower complication rate in the spinal anesthesia group as compared with the general anesthesia group (11.63% versus 15.28%; p = 0.0152) (see Appendix).
Multivariate logistic regression analysis with propensity score inclusion was used to adjust for potential confounders in a separate model. After propensity score adjustment, all differences in preoperative patient characteristics lost significance (p > 0.05), indicating a successful reduction in selection bias (Table I). After adjustment, the overall likelihood of complications was significantly greater in the general anesthesia group (p = 0.043; odds ratio [OR], 1.129; 95% confidence interval [CI], 1.004 to 1.269). The variables of age, sex, race, elevated serum creatinine, ASA class, operative time, and anesthetic choice were all independent risk factors for short-term complications after primary total knee arthroplasty (Table III). With use of an age of fifty to fifty-nine years as the reference group, an age of seventy to seventy-nine years (p < 0.0001; OR, 1.531; 95% CI, 1.263 to 1.856) and of eighty years or more (p < 0.0001; OR, 2.173; 95% CI, 1.725 to 2.737) predicted a higher complication rate. With use of male sex as a reference group, female sex (p = 0.0141; OR, 1.176; 95% CI, 1.033 to 1.338) predicted a higher complication rate. With use of white race as a reference group, black race (p < 0.0001; OR, 1.678; 95% CI, 1.346 to 2.092) predicted a higher complication rate. A serum creatinine value of >1.2 mg/dL (p < 0.0001; OR, 1.474; 95% CI, 1.243 to 1.748) predicted a higher complication rate. ASA class 3 or 4 (p = 0.0112; OR, 1.204; 95% CI, 1.060 to 1.367) predicted a higher complication rate. Finally, longer operative time, when used as a continuous variable, demonstrated a higher complication rate (p = 0.0007; OR, 1.003; 95% CI, 1.001 to 1.004).
The purpose of the present study was to determine the effect of anesthetic choice on short-term complication rates following total knee arthroplasty. We noted small differences between the spinal and general anesthesia groups in terms of the unadjusted rates of superficial surgical site infection, intraoperative blood transfusions, operative time, and hospital stay. After adjustment for confounders with use of a multivariate logistic regression and propensity analysis, the total risk of complications remained higher in the general anesthesia group. Age, sex, race, elevated serum creatinine, ASA class, operative time, and anesthetic choice were all independent risk factors for short-term complications after total knee arthroplasty. The overall differences between spinal and general anesthesia were <1% in many comparisons, and the clinical importance of these differences is likely low. The observed differences were greatest in patients with the highest number of medical comorbidities, indicating that patients with more comorbidities may be the most likely to benefit from spinal anesthesia.
These results are consistent with those of multiple previous studies that have shown a benefit of neuraxial anesthesia relative to general anesthesia in terms of the risk of complications in orthopaedic patients2-8. However, the majority of those previous studies compared epidural with general anesthesia, and there have been few trials directly comparing spinal anesthesia with general anesthesia in patients managed with total knee arthroplasty5. Thus, the present study should provide additional support for spinal anesthesia as an acceptable method of neuraxial anesthesia for patients undergoing total knee arthroplasty.
The propensity analysis divided patients into quintiles on the basis of the number of preoperative comorbidities. In the subsequent matched analysis, patients in the highest quintile of preoperative comorbidities who received general anesthesia had significantly more complications compared with patients who received spinal anesthesia. There was no difference between groups in the lower quintiles. Thus, patients with multiple comorbidities are the most likely to benefit from spinal anesthesia, and this modality should strongly be considered for the patient with comorbidities.
In the present study, age, sex, race, elevated serum creatinine, ASA class, operative time, and anesthetic choice were all independent risk factors for an increased thirty-day complication rate. These results are consistent with previous studies showing that black race, sex, general anesthesia, and older age are significant predictors of increased complication rates23-25. To our knowledge, the effects of serum creatinine, ASA class, and operative time have not been widely reported, and our study should make an argument for inclusion of those variables as independent risk factors for complications.
Most of the information on risk factors in patients undergoing total knee arthroplasty has come from large Medicare databases. Some of our reported risk factors, including older age, increased number of medical comorbidities, operative time, and black race are consistent with those studies26-29. However, congestive heart failure and chronic obstructive pulmonary disease26, low income28, hospital operative volume29, and male sex27 have been reported as risk factors in studies based on information from Medicare databases but were not independent risk factors in our analysis. Data on income and hospital volume are not available in the NSQIP database, and thus we cannot comment on the impact of socioeconomic status or hospital characteristics in our patient population. Other discrepancies may occur because most Medicare database studies identify risk on the basis of singular outcomes such as mortality or readmission, whereas the NSQIP database includes information on multiple short-term complications. Differences in data-collection methods also may account for discrepancies. For example, the Medicare database is based on coded billing claims. In contrast, the NSQIP database utilizes clinical reviewers who prospectively collect complications data and whose reviews are internally validated for accuracy. In addition, Medicare databases primarily represent patients over sixty-five years of age, whereas the NSQIP database includes all patients over eighteen years of age.
The present study had several weaknesses. First, the NSQIP database was not specifically designed to build predictive models of complications or mortality. Thus, our data are primarily representative of associations between risk factors and the subsequent operative complications. Furthermore, it is not an orthopaedic-specific database, and orthopaedic cases represent only about 7% of the total cases within the database. Consequently, orthopaedic-specific short-term outcomes such as postoperative pain and knee motion are unknown. Second, reporting of complications is limited to thirty days. As some complications such as deep-vein thrombosis and pulmonary embolism occur after thirty days, we cannot comment on the long-term differences between anesthetic methods. Third, spinal anesthesia is likely to be more prevalent at high-volume centers that have substantial expertise in total knee arthroplasty. The complication rates would be expected to be lower at these high-volume centers29, and it is not known what percentage of patients from each group received their care at a high-volume facility. Fourth, although the data were collected prospectively, our patients were not randomized, thus introducing the possibility of bias into the data analysis. We attempted to correct for this with two separate models involving the use of propensity analysis, and previous authors have shown that a propensity analysis with properly matched cohorts allows for comparisons that approach the statistical validity of a randomized study22. Furthermore, to our knowledge, this is the largest study to investigate differences in complications between general and spinal anesthesia in patients undergoing total knee arthroplasty, and the data were collected from 258 hospitals around the United States. Thus, these results are highly generalizable, and we believe that the large numbers of patients outweigh the downsides associated with database-based analyses.
In summary, these results indicate that patients undergoing total knee arthroplasty who receive spinal anesthesia have a significantly decreased risk of complications compared with patients who receive general anesthesia. The large number of patients in the present study provided sufficient power to find very small differences between groups. In many cases, the observed differences were <1%, and the clinical importance of such a small gap is not clear. Therefore, it is important to recognize that the differences in complication rates between spinal and general anesthesia are small and that general anesthesia remains a reasonable choice for many patients undergoing total knee arthroplasty.
The differences are greatest in patients with multiple medical comorbidities, and spinal anesthesia may have a greater benefit in that group. Furthermore, we have identified independent risk factors for complications after total knee arthroplasty, including patient age, sex, race, elevated serum creatinine, ASA class, operative time, and anesthetic choice. Our data should provide further support for consideration of those factors when determining operative risks in total knee arthroplasty.
Figures showing the cohort propensity score population distribution and propensity-matched quintiles for thirty-day complications rates between spinal anesthesia and general anesthesia and additional technical information on propensity score modeling are available with the online version of this article as a data supplement at jbjs.org.
Investigation performed at the University of Iowa Hospitals and Clinics, Iowa City, Iowa
Disclaimer: The American College of Surgeons National Surgical Quality Improvement Program and the hospitals participating in the ACS NSQIP are the source of the data used herein; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors.
This article was chosen to appear electronically on December 26, 2012 in advance of publication in a regularly scheduled issue.
1. Center for Disease Control and Prevention. National hospital discharge survey 1998-2005. http://http://www.cdc.gov
/nchs/nhds.htm. Accessed 2011 Oct 25.
2. Davis FM Laurenson VG Gillespie WJ Wells JE Foate J Newman E. Deep vein thrombosis after total hip replacement. A comparison between spinal and general anaesthesia. J Bone Joint Surg Br. 1989;71(2):181–5.
3. Chakladar A White SM. Cost estimates of spinal versus general anaesthesia for fractured neck of femur surgery. Anaesthesia. 2010;65(8):810–4.
4. Rashiq S Finegan BA. The effect of spinal anesthesia on blood transfusion rate in total joint arthroplasty. Can J Surg. 2006;49(6):391–6.
5. Hu S Zhang ZY Hua YQ Li J Cai ZD. A comparison of regional and general anaesthesia for total replacement of the hip or knee: a meta-analysis. J Bone Joint Surg Br. 2009;91(7):935–42.
6. Chang CC Lin HC Lin HW Lin HC. Anesthetic management and surgical site infections in total hip or knee replacement: a population-based study. Anesthesiology. 2010;113(2):279–84.
7. Maurer SG Chen AL Hiebert R Pereira GC Di Cesare PE. Comparison of outcomes of using spinal versus general anesthesia in total hip arthroplasty. Am J Orthop (Belle Mead NJ). 2007;36(7):E101–6.
8. Gonano C Leitgeb U Sitzwohl C Ihra G Weinstabl C Kettner SC. Spinal versus general anesthesia for orthopedic surgery: anesthesia drug and supply costs. Anesth Analg. 2006;102(2):524–9.
9. Planès A Vochelle N Fagola M Feret J Bellaud M. Prevention of deep vein thrombosis after total hip replacement. The effect of low-molecular-weight heparin with spinal and general anaesthesia. J Bone Joint Surg Br. 1991;73(3):418–22.
10. Nielson WR Gelb AW Casey JE Penny FJ Merchant RN Manninen PH. Long-term cognitive and social sequelae of general versus regional anesthesia during arthroplasty in the elderly. Anesthesiology. 1990;73(6):1103–9.
11. Horlocker TT. Complications of regional anesthesia and acute pain management. Anesthesiol Clin. 2011;29(2):257–78.
12. Schoenfeld AJ Ochoa LM Bader JO Belmont PJ Jr. Risk factors for immediate postoperative complications and mortality following spine surgery: a study of 3475 patients from the National Surgical Quality Improvement Program. J Bone Joint Surg Am. 2011;93(17):1577–82.
13. Belmont PJ Jr Davey S Orr JD Ochoa LM Bader JO Schoenfeld AJ. Risk factors for 30-day postoperative complications and mortality after below-knee amputation: a study of 2,911 patients from the national surgical quality improvement program. J Am Coll Surg. 2011;213(3):370–8.
14. Schilling PL Hallstrom BR Birkmeyer JD Carpenter JE. Prioritizing perioperative quality improvement in orthopaedic surgery. J Bone Joint Surg Am. 2010;92(9):1884–9.
15. Greenblatt DY Rajamanickam V Pugely AJ Heise CP Foley EF Kennedy GD. Short-term outcomes after laparoscopic-assisted proctectomy for rectal cancer: results from the ACS NSQIP. J Am Coll Surg. 2011;212(5):844–54.
16. Kelly KJ Greenblatt DY Wan Y Rettammel RJ Winslow E Cho CS Weber SM. Risk stratification for distal pancreatectomy utilizing ACS-NSQIP: preoperative factors predict morbidity and mortality. J Gastrointest Surg. 2011;15(2):250–9, discussion :259-61. 2010 Dec 15 [Epub ahead of print].
17. Suleiman LI Ortega G Ong’uti SK Gonzalez DO Tran DD Onyike A Turner PL Fullum TM. Does BMI affect perioperative complications following total knee and hip arthroplasty? J Surg Res. 2012 May 1;174(1):7–11. Epub 2011 Jun 25.
18. Paquette IM Swenson BR Kwaan MR Mellgren AF Madoff RD. Thirty-day outcomes in patients treated with en bloc colectomy and pancreatectomy for locally advanced carcinoma of the colon. J Gastrointest Surg. 2012 Mar;16(3):581–6. Epub 2011 Sep 29.
19. Davenport DL Xenos ES. Early outcomes and risk factors in venous thrombectomy: an analysis of the American College of Surgeons NSQIP dataset. Vasc Endovascular Surg. 2011;45(4):325–8.
20. Page AJ Pollock JD Perez S Davis SS Lin E Sweeney JF. Laparoscopic versus open appendectomy: an analysis of outcomes in 17,199 patients using ACS/NSQIP. J Gastrointest Surg. 2010;14(12):1955–62.
21. American College of Surgeons National Surgical Quality improvement program. Program specifics: ACS NSQIP data collection overview. http://acsnsqip.org/main/programspecs/program_data_collection.jsp. Accessed 2011 Oct 25.
22. D’Agostino RB Jr. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med. 1998;17(19):2265–81.
23. Santaguida PL Hawker GA Hudak PL Glazier R Mahomed NN Kreder HJ Coyte PC Wright JG. Patient characteristics affecting the prognosis of total hip and knee joint arthroplasty: a systematic review. Can J Surg. 2008;51(6):428–36.
24. Ibrahim SA Stone RA Han X Cohen P Fine MJ Henderson WG Khuri SF Kwoh CK. Racial/ethnic differences in surgical outcomes in veterans following knee or hip arthroplasty. Arthritis Rheum. 2005;52(10):3143–51.
25. Higuera CA Elsharkawy K Klika AK Brocone M Barsoum WK. 2010 Mid-America Orthopaedic Association Physician in Training Award: predictors of early adverse outcomes after knee and hip arthroplasty in geriatric patients. Clin Orthop Relat Res. 2011;469(5):1391–400.
26. Huddleston JI Maloney WJ Wang Y Verzier N Hunt DR Herndon JH. Adverse events after total knee arthroplasty: a national Medicare study. J Arthroplasty. 2009;24(6)(Suppl):95-100.
27. Kurtz SM Ong KL Lau E Bozic KJ Berry D Parvizi J. Prosthetic joint infection risk after TKA in the Medicare population. Clin Orthop Relat Res. 2010;468(1):52–6.
28. Mahomed NN Barrett J Katz JN Baron JA Wright J Losina E. Epidemiology of total knee replacement in the United States Medicare population. J Bone Joint Surg Am. 2005;87(6):1222–8.
29. Katz JN Barrett J Mahomed NN Baron JA Wright RJ Losina E. Association between hospital and surgeon procedure volume and the outcomes of total knee replacement. J Bone Joint Surg Am. 2004;86-A(9):1909–16.
Disclosure: None of the authors received payments or services, either directly or indirectly (i.e., via his or her institution), from a third party in support of any aspect of this work. One or more of the authors, or his or her institution, has had a financial relationship, in the thirty-six months prior to submission of this work, with an entity in the biomedical arena that could be perceived to influence or have the potential to influence what is written in this work. No author has had any other relationships, or has engaged in any other activities, that could be perceived to influence or have the potential to influence what is written in this work. The complete Disclosures of Potential Conflicts of Interest submitted by authors are always provided with the online version of the article.