Application of the American College of Cardiology (ACC)/American Heart Association (AHA) guidelines to preoperative cardiac evaluation (1) requires information on the risk of the proposed surgery as well as the urgency/emergency of the operation, the recency of cardiac workup and therapy, preexisting comorbidities, and functional status. Appropriate implementation of the guidelines has been shown to lead to a selective preoperative cardiac testing with cost savings, without adversely affecting perioperative cardiac complication rates (2). Whereas several previous studies on major lower extremity amputations found perioperative 30-day mortality rates in the range of 8%–23% (3–6), the incidence of cardiac events was not reported. To apply the ACC/AHA guidelines to patients undergoing major amputations, it is necessary to define the incidence of cardiac events, which, according to the ACC/AHA, are defined as either nonfatal myocardial infarction (MI) or death from a cardiac cause. Surgeries that are associated with ≥5% cardiac event rates are then classified as high risk, whereas those with 1%–5% cardiac event rates are intermediate risk and those with <1% cardiac event rates are low risk. Accordingly, in the present study, we conducted a retrospective review of all major amputations over the last decade at our university hospital to define the perioperative mortality and cardiac event rates after above-knee and below-knee amputations (AKA, BKA). Correct risk classification of these surgeries would be a first step toward any risk reduction strategies.
Second, whereas diabetes mellitus (DM) is listed as an intermediate clinical predictor of perioperative cardiac risk in the ACC/AHA guidelines (1), there are conflicting data in the recent literature as to whether DM is truly an independent predictor of perioperative cardiac events in major noncardiac surgery. In a retrospective review of >4000 patients undergoing major noncardiac surgeries, Lee et al. (7) identified 6 independent predictors of cardiac events, which included insulin treatment for DM, history of coronary artery disease (CAD), congestive heart failure (CHF), cerebrovascular disease, high-risk surgery, and serum creatinine >2 mg/dL. More recently, Boersma et al. (8) did not find that DM was an independent risk factor of cardiac events in major vascular surgery, although they corroborated that history of angina, MI, CHF, and cerebrovascular accidents (CVA) were important predictors, as in the study by Lee et al. Therefore, in this study, we have also examined whether DM is a significant predictor of cardiac events after major amputations. We performed the analysis both perioperatively and for 3 and 10 years postoperatively.
After obtaining IRB approval, the vascular surgical quality assurance database at Beth Israel Deaconess Medical Center was reviewed to identify those patients who had undergone AKA or BKA between 1990 and May 2001. Revisions of amputations that occurred during the same hospitalization as the initial amputation were not considered to be independent operations and were not counted. However, an AKA that occurred more than 30 days after a previous BKA in the same patient was considered a separate operation. Similarly, a contralateral limb amputation was counted as a separate operation. A dedicated database technician had completed data entry as care was provided. The database was retrospectively reviewed for demographic variables, comorbidities, date and nature of surgical procedure, and occurrence of complications (MI, CHF, and death). A postoperative electrocardiogram (ECG) was obtained selectively in all patients with a history of DM, CAD, CHF, or arrhythmias, or those who had any intraoperative ECG changes suggestive of ischemia. The ECG was initially read by the anesthesiologist and surgical team and, if any new changes were found, confirmed by a cardiologist. New ECG changes were defined (9) as horizontal or downsloping ST segment depression of at least 0.1 mV or ST elevation of at least 0.15 mV; ECG changes led to 3 serial cardiac isoenzymes and ECGs every 8 h. Postoperative MI was defined by creatinine kinase MB fraction >5% and/or troponin-I >2.0 ng/mL in the presence of appropriate ECG changes. Signs of pulmonary edema on chest radiograph, in conjunction with the appropriate clinical symptoms/signs such as orthopnea and pulmonary rales, confirmed a diagnosis of CHF. Significant arrhythmias were defined as new-onset atrial fibrillation, supraventricular tachycardia, or ventricular tachyarrhythmias requiring intervention. Perioperative mortality was defined as death within 30 days of the operation. Causes of mortality were identified as either cardiac (post-MI, arrhythmias, or heart failure) or noncardiac. We collected data on the long-term survival using the social security database available at www.ancestry.com.
Continuous variables were analyzed by Student’s t-test and proportions were analyzed by Yates corrected χ2 test. In order to identify predictors of perioperative mortality and perioperative cardiac events, univariate analyses were initially conducted on the demographic variables, comorbidities, site of amputation (AKA versus BKA), and indications for surgeries. A dummy variable was introduced to account for a possible confounding influence of time: the study period was divided into roughly equal thirds of 1990–93, 1994–97, and 1998–2001. These periods were then each assigned a dummy variable value of 1, 2, and 3, respectively. Variables that predicted outcome with a P < 0.10 on the univariate analyses were then used to perform a multivariate logistic regression analysis, except that age >64 yr, gender, DM, and the dummy variable were forced into the multivariate analysis regardless of the result of the initial univariate analysis.
By checking for long-term survival in May 2004, the minimum duration of follow-up was 3 yr for all patients and the maximum possible was 14 yr for those operated in the 1990s. Because survival beyond 3 yr was censored for those operated on in the recent years, we analyzed long-term survival to 3 yr and beyond 3 yr by Kaplan-Meier analysis. Kaplan-Meier curves were initially compared by log-rank test. To account for the effect of covariates on survival, Cox regression analysis was then performed. P < 0.05 was considered significant. Statistical analysis was done by using STATA version 8 (StataCorp LP, College Station, TX).
There were 954 major amputations (720 BKA, 234 AKA) in 762 patients, with a male: female ratio of 436:326 (57%: 43%). Age at the time of the operation was 66.1 ± 14.0 yr (mean ± sd). Eighty-eight percent were Caucasians, 9% were African-Americans, and 3% were from other races. Nearly 25% of the patients had 2 or more amputations; 19.5% had two, 4% had three, and 0.5% had four.
As for the intermediate clinical predictors (ICP) of the ACC/AHA, 82% had DM (63% were insulin-treated), 37% had a history of MI, 10% had stable angina, 36% had a history of CHF, and 38% had chronic renal insufficiency (CRI) with serum creatinine >2 mg/dL, with 18% on dialysis (14% on hemodialysis and 4% on peritoneal dialysis). Ninety-one percent had at least one ICP, 64% had at least 2 ICPs, and 35% had at least 3. As for other comorbidities, 73% were smokers, 66% had hypertension, 21% had a history of CVA, and 51% had hypercholesterolemia. Twenty-four percent had undergone coronary artery bypass grafting, 7% had history of percutaneous transluminal coronary angioplasty, and 4% had kidney transplants. Indications for amputations in our population were listed as graft failure or occlusion in 222 patients, rest pain or disabling claudication in 416, nonhealing ulcer in 447, gangrene in 689, and infection in 402, including 22 with abscesses and 46 with systemic sepsis. These figures add up to more than the total number of amputations because some patients had more than one of these indications.
Overall, the incidence of perioperative MI, CHF, and significant arrhythmias were 1.9%, 3.8%, and 2.5%, respectively. Other significant morbidities included pneumonia in 3.0%, acute renal failure in 2.1%, and CVA in 0.5%. Perioperative 30-day mortality was 7.4% (71 deaths after 954 operations). Twenty-eight of these deaths were attributable to definable cardiac causes (MI, CHF, or arrhythmias). Sepsis accounted for 8 deaths, renal failure for 6 deaths, CVA for 3 deaths, and respiratory failure for 2 deaths. Liver failure, abdominal bleeding, aortic dissection, pulmonary embolism, pneumonia, and colovaginal fistula accounted for one death each. In the remaining 18 patients, the causes of the deaths could not be identified, either because no clearly identifiable cause appeared responsible for the death from record review or because the deaths occurred after discharge from the hospital, though within 30 days of the operation. Thus, cardiac deaths could have ranged from 2.9% (28 of 954) to 4.8% (46 of 954), depending on how many of the deaths with unidentified causes were attributable to cardiac causes. The perioperative cardiac event rate, including cardiac deaths and nonfatal MI, was at least 3.6% (34 of 954) but could have been as much as 5.5% (52 of 954), depending on the nature of the post-discharge deaths. The perioperative cardiac event rate for AKA was at least 6.8% (16 of 234, 16 deaths from identified cardiac causes) but could have been as much as 11.5% (27 of 234), depending on the nature of the postdischarge deaths. For BKA, the perioperative cardiac event rate was at least 2.6% (19 of 720, 12 cardiac deaths and 7 nonfatal MI) but could have been as much as 3.6% (26 of 720), depending on the nature of the postdischarge deaths.
On univariate analyses, variables that predicted 30-day mortality with P < 0.10 were gender, age >64 yr, history of CHF, CAD, MI, CVA, renal insufficiency, requirement for dialysis (either hemodialysis or peritoneal dialysis), and site of amputation. These variables, as well as history of DM and the time dummy variable, were entered into a multivariate logistic analysis of perioperative 30-day mortality. Only 2 variables were found to predict 30-day mortality significantly: renal insufficiency (P = 0.019) and site of amputation (P < 0.001) (Table 1). Perioperative 30-day mortality was 17.5% after AKA and 4.2% after BKA. The 30-day mortality was 10.9% with a history of renal insufficiency and 5.3% without. Although history of DM did not predict perioperative 30-day mortality (P = 0.43), it was notable that the mortality was 9.4% in nondiabetics, 10.2% among diabetics taking oral hypoglycemics, and 6.1% among those taking insulin.
On univariate analyses, variables that predicted perioperative death, MI, or CHF with P < 0.10 were history of CAD, CHF, MI, CVA, renal insufficiency, requirement for dialysis, site of amputation, and the dummy variable for time. These variables, as well as gender, age >64 yr, and DM, were entered into a multivariate analysis to predict perioperative death, MI, or CHF. Four variables were found to predict perioperative death, MI, or CHF: history of MI (P = 0.031), renal insufficiency (P = 0.033), site of amputation (P < 0.001), and the dummy variable (P = 0.008) (Table 2). There was a suggestion that a fifth variable, history of CVA, could also predict perioperative death, MI, or CHF (P = 0.069). Perioperative death, MI, or CHF rates were 14.8% for patients with history of MI and 7.1% for those without, 14.0% for patients with renal insufficiency and 7.6% for those without, 15.5% for those with previous CVA and 8.7% for those without, 20.5% for those who underwent AKA and 6.7% for BKA patients, and 13.9% in 1990–93, 8.6% in 1994–97, and 8.4% in 1998–2001. Again, DM failed to predict perioperative death, MI, or CHF (P = 0.88). It was notable, however, that perioperative death, MI, or CHF was 10.5% in nondiabetics, 13.2% in diabetics taking oral hypoglycemics, and 9.1% in those taking insulin.
Three-year survival was significantly different between diabetic amputees and nondiabetic amputees by log-rank test (P = 0.021, Fig. 1). However, when the two survival curves were controlled for coexisting covariates by Cox regression analysis, the effect of diabetes on 3-yr survival became nonsignificant (Table 3). Significant predictors of 3-yr mortality were age >64, history of CHF, requirement for hemodialysis, and the site of amputation (AKA).
Long-term survival was significantly different between diabetic amputees and nondiabetic amputees by log-rank test (P < 0.001) (Fig. 2). Even when the two survival curves were controlled for coexisting covariates by Cox regression analysis, the effect of diabetics on long-term survival remained significant (Table 4). Other significant predictors of long-term mortality were age >64 yr, history of CHF, history of MI, history of renal insufficiency, requirement for dialysis, and the site of amputation (AKA) as well as DM.
The site of amputation (AKA versus BKA) affected long-term survival significantly. Long-term survival was 62.1% ± 3.4% at 1 yr, 47.7% ± 3.5% at 2 yr, 31.5% ± 3.5% at 5 yr, and 19.0% ± 3.9% at 10 yr for AKA patients and 78.2% ± 1.6% at 1 yr, 67.1% ± 1.8% at 2 yr, 47.9% ± 2.0% at 5 yr, and 28.5% ± 2.7% at 10 yr for BKA patients (Figs. 3 and 4). The median survival for AKA was significantly less (20 mo) compared with BKA (52 mo) (P < 0.001).
The most significant findings in the present study are that 1) the perioperative adverse cardiac event (cardiac death and nonfatal MI) rate was at least 6.8% in AKA and at most 3.6% in BKA patients, 2) DM was not a significant predictor of perioperative cardiac events or death or postoperative 3-year survival but was a significant predictor of longer-term (10-year) survival after major amputations, and 3) the site of amputation was a significant predictor of perioperative 30-day mortality, perioperative death, MI, or CHF, 3-year survival, and 10-year survival.
Our findings of frequent perioperative mortality of 7.4% for AKA and BKA patients are consistent with the findings of earlier reports (3–6). Unlike previous reports, we sought to distinguish deaths from cardiac causes and other deaths, so that we could determine the cardiac event rate as defined by the ACC/AHA guidelines (1). Even though the causes of death after discharge from the hospital but within 30 days of the surgery were often unknown, our data indicate that the perioperative cardiac event rate was at least 6.8% for AKA and at most 3.6% for BKA. Thus for risk stratification per the ACC/AHA guidelines, BKA should be triaged as intermediate-risk surgeries for cardiac risk and AKA as high-risk surgeries. Strict adherence to the ACC/AHA guidelines would thus require that for BKA, the patients should obtain preoperative cardiac evaluation when they have both poor functional status and an intermediate clinical predictor, whereas for AKA, the patients would warrant preoperative cardiac workup for either poor functional status or an intermediate clinical predictor. As functional status is not easily assessed in the majority of these patients, preoperative cardiac workup may often be obtained in the presence of an intermediate clinical predictor alone. Surviving patients also had frequent cardiac morbidities (1.9% MI, 3.8% CHF, and significant arrhythmias 2.5%) in the perioperative period, suggesting the need for preoperative cardiac evaluation and continued perioperative monitoring of the cardiovascular system.
As defined by the ACC/AHA guidelines on preoperative cardiac evaluation, our patients had a number of ICPs of increased perioperative cardiovascular risk. In our cohort, the incidence of ICPs was frequent, with DM 82%, history of MI 37%, history of CHF 36%, and history of CRI 37%. Ninety-one percent of the patients had one of the ICPs and 64% of patients had two or more. Thus a majority of our patients would have warranted a preoperative cardiac evaluation.
In a retrospective review of 4315 patients undergoing major noncardiac surgery, Lee et al. (7) found 6 independent predictors of cardiac complications: namely, high risk surgery, history of CAD, history of CHF, history of cerebrovascular disease, preoperative treatment with insulin, and preoperative creatinine >2.0 mg/dL. The presence of 0, 1, 2, or ≥ 3 of these risk factors was associated with 0.4, 0.9, 7, and 11% risks, respectively, of cardiac events. Likewise, Boersma et al.’s (8) review of vascular surgical patients showed that the presence of multiple risk factors led to a worse cardiac outcome than the presence of a single risk factor. A strategy one may adopt in patients whose functional capacity is difficult to assess may be to obtain cardiac evaluation in the presence of 2 or more ICPs from the ACC/AHA guidelines. Even such a strategy would have led to cardiac workup in approximately two thirds of our amputation patients. Future studies should address whether identifying patients at risk for perioperative myocardial ischemia and applying risk reduction strategies might have reduced perioperative cardiac events.
β-adrenergic blockers decrease perioperative and long-term cardiac morbidity and mortality (10,11). In addition, the use of statins has been suggested to reduce perioperative mortality in patients undergoing major vascular surgery (12). The efficacy of β-adrenergic blockers or statins in patients presenting for amputations has not been studied. We did not have data on how many of our patients received adequate β-adrenergic blockade or statins in the perioperative period and could not ascertain whether these measures affected cardiac outcome. Because at least 39% (28 of 71) of the perioperative deaths in our patients were from cardiac causes, future studies on the efficacy of β-blockers or statins in these patients may be indicated. In addition, because a large percentage of the deaths after major amputations were from noncardiac causes, a multifaceted approach to mortality reduction, such as better infection control and pulmonary toilet, may be indicated.
In our amputation patients, DM was not a significant predictor of perioperative mortality or cardiac events or postoperative 3-year survival but was a significant predictor of longer-term survival. Although Roghi et al. (13) showed that in the “intermediate-risk” group of vascular patients, DM was a significant predictor of adverse cardiac events, Hamdan et al. (14), in their study on 6565 major vascular operations, and Axelrod et al. (15) in their study on more than 10,000 vascular surgeries, have shown DM was not an independent risk factor. Several factors may explain this difference. Silent MIs can be missed in diabetics in the perioperative period and thus underestimated. It is unknown if diabetics receive a better perioperative cardiac workup and thereby better preparation. Importantly, duration of DM and the quality of serum glucose control (e.g., as assessed by measurement of HbA1C) might expose any differences, if any. Finally, it is possible that the mortality and morbidity among diabetics may be different between those treated with insulin and those taking oral hypoglycemics. Sulfonylureas, which are used commonly as oral hypoglycemic drugs, block adenosine triphosphate-sensitive potassium channels, which are believed to be involved in ischemic preconditioning of the myocardium (16). Future studies examining the effect of diabetes on cardiovascular outcome should examine not only whether the patient has diabetes but also the quality of diabetic control (long-term HbA1C and acutely, as measured by serum glucose) and the treatment modality.
Whether DM should be considered a significant predictor of perioperative cardiac complications has an important practical implication on preoperative cardiac evaluation. In our patients, disregarding DM as an ICP would have resulted in only 68.8% of the patients with at least one ICP and 37.5% with at least 2 ICPs. Thus by disregarding DM as a significant predictor of perioperative cardiac outcome, we might have required a preoperative cardiac workup in 30% fewer of our patients.
Surgical decisions for amputations may often be based on the unproven presumption that amputations represent a simpler and therefore less risky alternative than vascular bypass procedures. In a Medicare population, peripheral vascular bypass surgeries had a 30-day mortality of 7.3% (17). Our data show that major amputations have a comparable 30-day mortality rate. Whereas 30-day mortality rates are comparable, long-term outlook for survival is significantly different between the two types of procedures. In a series of 1498 peripheral vascular bypass patients, Landry et al. (18) reported survival rates of 72.4% at 5 years, 67.8% at 7 years, and 53.4% at 10 years. These figures compare favorably to the survival rates of amputees. Several authors have suggested (19–21) that successful crural or pedal bypass surgeries, as well as percutaneous vascular dilation or chemical lumbar sympathectomy, may be able to reduce the amputation rates. Significant differences in the outlook for long-term survival between amputees and nonamputees would suggest that a vascular bypass should be attempted whenever possible before proceeding with amputations.
Our findings of poor long-term survival in both AKA and BKA patients are consistent with previous studies of amputees showing similarly poor long-term survival. A Finnish study of 705 patients in 1984–85 (22) showed that after AKA, 54% lived longer than 1 year, 36% more than 2 years, 18% more than 5 years, and 8% more than 10 years. The corresponding figures for BKA were 70%, 53%, 21%, and 4%. Likewise, a Scottish study of 1710 patients from 1965–89 reported a median survival of 4 years 9 months after BKA and 4 years 3 months after AKA (23), whereas another Scottish study of 2759 patients from 1989–93 found a mean survival of 2 years (24). In the Scottish study of 1710 patients, DM was not a significant predictor of amputee survival. In our study, DM was not a significant predictor of 3-year survival but was associated with poor long-term survival beyond 3 years. In the general population, the average life expectancy is decreased to 15 years in diabetics (25).
There are several limitations of our retrospective data analysis. First, we did not have data on how and how well each of the comorbidities was being treated at the time of the surgery. In particular, we had no data on acute or long-term control of DM in our patients. Information of treatment and diabetic control might have shown the effect of poor diabetic control on perioperative or long-term cardiac events and death. Second, our long-term follow-up of the patients was complete for up to 3 years but incomplete beyond 3 years for those patients operated on in recent years. The Kaplan-Meier survival analysis takes into account data censored as a result of a fixed study period and uses the fact that although the exact time of the outcome event (i.e., death) may not be known for some patients, the fact that the event did not precede the censoring time is a piece of useful information. Nevertheless, interpretation of the survival data beyond 3 years should be performed with caution. Third, our approach to the diagnosis of perioperative MI would miss clinically silent MIs without hemodynamic consequences. Because there might have been some silent MIs, our perioperative MI rate might have been underestimated.
In summary, AKA and BKA are associated with frequent perioperative mortalities independent of the presence of diabetes. Based on cardiac event rates as defined by the ACC/AHA, AKA should be considered a high-risk surgery, whereas BKA should be considered an intermediate-risk surgery and be triaged as such for the purpose of preoperative cardiac evaluation. DM was not a significant predictor of either perioperative adverse cardiac events or 3-year survival after major amputations. Further studies are needed to determine whether different therapeutic approaches to these patients might alter their morbidity and mortality.
1. Eagle KA, Berger PB, Calkins H, et al. ACC/AHA Guideline Update for Perioperative Cardiovascular Evaluation for Noncardiac Surgery—Executive summary: A report of the ACC/AHA task force on practice guidelines (Committee to Update the 1996 Guidelines on Perioperative Cardiovascular evaluation for noncardiac surgery). J Am Coll Cardiol 2002;39:542–53.
2. Froehlich JB, Karavite D, Russman P, et al. American College of Cardiology/American Heart Association preoperative assessment guidelines reduce resource utilization before aortic surgery. J Vasc Surg 2002;36:758–63.
3. Kald A, Carlsson R, Nilsson E. Major amputation in a defined population: Incidence, mortality and results of treatment. Br J Surg 1989;76:308–10.
4. Kazmers A, Perkins A, Jacobs L. Major lower extremity amputation in veterans’ affairs medical centers. Ann Vasc Surg 2000;14:216–22.
5. Peng CW, Tan SG. Perioperative and rehabilitative outcomes after amputation for ischemic leg gangrene. Ann Acad Med 2000;29:168–72.
6. Toursarkissian B, Shireman PK, Harrison A, et al. Major lower-extremity amputation: Contemporary experience in single veterans affairs institution. Am Surg 2002;68:606–10.
7. Lee TH, Marcantonio ER, Mangione CM, et al. Derivation and prospective validation of a simple index for prediction of cardiac risk of major noncardiac surgery. Circulation 1999;100:1043–9.
8. Boersma E, Poldermans D, Bax JJ, et al. Predictors of cardiac events after major vascular surgery. JAMA 2001;285:1865–73.
9. Slogoff S, Keats AS. Does perioperative myocardial ischemia lead to postoperative myocardial infarction? Anesthesiology 1985;62:107–14.
10. Poldermans D, Boersma E, Bax JJ, et al. Bisoprolol reduces cardiac death and myocardial infarction in high-risk patients as long as 2 years after successful major vascular surgery. Eur Heart J 2001;22:1353–8.
11. Mangano DT, Layug EL, Wallace A, et al. Effect of atenolol on mortality and cardiovascular morbidity after noncardiac surgery: Multicenter Study of Perioperative Ischemia Research Group. N Engl J Med 1996;335:1713–20.
12. Poldermans D, Bax JJ, Kertai MD, et al. Statins are associated with a reduced incidence of perioperative mortality in patients undergoing major noncardiac vascular surgery. Circulation 2003;107:1848–51.
13. Roghi A, Palmieri B, Crivellaro W, et al. Relationship of unrecognized myocardial infarction, diabetes mellitus and type of surgery to postoperative cardiac outcomes in vascular surgery. Eur J Vasc Endovasc Surg 2001;21:9–16.
14. Hamdan AD, Saltzberg SS, Sheahan M, et al. Lack of association of diabetes with increased postoperative mortality and cardiac morbidity: Results of 6565 major vascular operations. Arch Surg 2002;137:417–21.
15. Axelrod DA, Upchurch GR, DeMonner S, et al. Perioperative cardiovascular risk stratification of patients with diabetes who undergo elective major vascular surgery. J Vasc Surg 2002;35:894–901.
16. Gu W, Pagel PS, Warltier DC, Kersten JR. Modifying cardiovascular risk in diabetes mellitus. Anesthesiology 2003;98:774–9.
17. Fleisher LA, Eagle KA, Shaffer T, Anderson GF. Perioperative and long-term mortality rates after major vascular surgery: The relationship to preoperative testing in the Medicare population. Anesth Analg 1999;89:849–55.
18. Landry GJ, Moneta GL, Taylor LM Jr., et al. Long-term outcome of revised lower-extremity bypass grafts. J Vasc Surg 2002;35:56–62.
19. Lindholt JS, Bovling S, Fasting H, Henneberg EW. Vascular surgery reduces the frequency of lower limb major amputations. Eur J Vasc Surg 1994;8:31–5.
20. Pedersen AE, Bornefeldt Olsen B, et al. Halving the number of leg amputations: The influence of infrapopliteal bypass. Eur J Vasc Surg 1994;8:26–30.
21. Johansen K, Burgess EM, Sorn R, et al. Improvement of amputation level by lower extremity revascularization. Surg Gynecol Obstet 1981;153:707–9.
22. Pohjolainen T, Alaranta H. Ten year survivals of Finnish lower limb amputees. Prosthet Orthot Int 1998;22:10–6.
23. Stewart CP, Jain AS, Ogston SA. Lower limb amputee survival. Prosthet Orthot Int 1992;16:11–8.
24. Pell J, Stonebridge J. Association between age and survival following major amputation. The Scottish Vascular Audit Group. Eur J Vasc Endovasc Surg 1999;17:166–9.
25. Kahn RC, for the Diabetes Research Working Group. Summary of the report and recommendations of the congressionally established Diabetes Research Working Group: A strategic plan for the 21st century. February 1999. Available at: http://www.jdfcure.org/drwg
. Accessed August 15, 2001.