Relation of Lower-Extremity Amputation to All-Cause and Cardiovascular Disease Mortality in American Indians: The Strong Heart Study

The Strong Heart Study (SHS) was initiated in 1988 to investigate CVD and its risk factors in American Indians (15). The design, methods, and laboratory techniques of the SHS have been reported previously (15–17). The SHS cohort consists of 4,549 participants aged 45–74 from Arizona, Oklahoma, and South and North Dakota who were seen at the baseline examination, which was conducted between 1989 and 1992. The second and third examinations were conducted between 1993 and 1995 and between 1997 and 1999, respectively. Nonparticipants were similar to participants in age, BMI, and self-reported frequency of diabetes and hypertension (17). The SHS was approved by the institutional review boards of participating institutions, and all participants provided informed consent. Surveillance for cause-specific mortality is complete through 31 December 2000.

At each of the three SHS examinations, trained examiners collected amputation data from direct observation of both legs. Data from this exam included presence/absence and cause of missing extremities. The location of missing extremities was coded by study personnel as toe, below the knee (BKA), or above the knee (AKA). Participants were asked to attribute each missing extremity to a specific factor. These attributions were coded as diabetes, trauma, congenital, other, and unknown. In this report, LEA refers to amputations that were caused by diabetes.

Participants were categorized as diabetic if they had a fasting glucose ≥126 mg/dl, reported use of hypoglycemic medications, or reported diabetes at the baseline examination.

Previous studies of LEA in American Indians and other populations provide a rationale for selection of potential LEA risk factors. These include age, glycemic control, smoking, hypertension, and male sex (5,6,9,10,14,18). Participants were considered hypertensive if they had a systolic blood pressure ≥140 mmHg, a diastolic blood pressure ≥90 mmHg, or if they were taking antihypertension medication. Smoking history was determined by questionnaire. Excessive alcohol consumption can cause or aggravate neuropathy. In this report, at-risk drinking was defined as a report of either ≥5 drinks on one occasion in the past month or ≥14 drinks in a typical week.

Methods for measurement of CVD risk factors, including BMI, waist circumference, ankle and arm blood pressures, and renal function have been described and laboratory methods have been published previously (15,16,19). Cholesterol, triglycerides, and fasting glucose were determined by enzymatic methods using a Hitachi chemistry analyzer and consistent, standardized reagents (Boehringer Mannheim Diagnostics, Indianapolis, IN). Fasting insulin and fibrinogen were measured by established methods (20,21). Urinary albumin excretion was estimated by the ratio of albumin (mg) to creatinine (g) in a morning urine sample. Microalbuminuria was defined as a ratio of urinary albumin (mg/ml) to creatinine (g/ml) of 30–299 mg/g, and macroalbuminuria was defined as a ratio of >300 mg/g.

An ankle brachial index (ABI) of <0.90 (“low ABI”) is 95% sensitive and 99% specific for angiographically documented peripheral arterial disease (22,23), and it has been used in previous work in the SHS (19). Although it is not a definitive assessment of medial arterial calcification, data from the SHS (24) show that ABI >1.40 (“high ABI”) is associated with a similar level of CVD risk as ABI <0.90. Baseline ABI was used to categorize three groups of individuals: those with “low” ABI (<0.90), those in the “normal” range of ABI (0.90 ≤ ABI ≤ 1.40), and those with “high” ABI (>1.40).

Deaths among cohort members were identified through tribal and Indian Health Service (IHS) hospital records and through direct contact with participant families (15,17). Death certificates were obtained from state health departments, and ICD-9 codes were applied centrally by a nosologist (25). The SHS Mortality Review Committee independently confirmed the cause of each study participant death by review of autopsy reports, medical records, and informant interviews, using published criteria to ascertain deaths due to fatal myocardial infarction, coronary heart disease, and stroke (25).

We used χ² and Student’s t tests to examine differences in proportions and means between categorical and continuous variables, respectively. Initial comparisons focused on differences in risk factors between participants with and without an LEA at the baseline examination. Subsequent analyses focused on differences in risk factors between participants with an LEA who did and did not experience a recurrent amputation.

Person-years were calculated from the date of the baseline examination to the date of a fatal event, the date of the last exam if the participant was lost to follow up (n = 7; 0.2%), or 31 December 2000. Sex-specific rates of all-cause and CVD mortality were calculated according to baseline LEA status, and crude rate ratios quantifying the effect of LEA on mortality were estimated. Cox regression models were used to determine whether LEA predicted all-cause or CVD mortality, adjusting for potential confounders. These analyses are presented for the whole sample and stratified by sex. We tested the proportional hazards assumption by plotting hazard curves for LEA against the log of time. No evidence was found that this assumption was violated.

Survival analysis was used to visualize survival patterns among diabetic participants with an LEA at specific sites (toe, BKA) compared with those without an LEA and individuals without diabetes. The log-rank test was used to evaluate differences in survivorship among these categories, and the Cox model was used to quantify adjusted hazard ratios (HRs) associated with LEA at different anatomical sites. We examined duration of diabetes in relation to the location of LEAs. Because diabetes duration was not normally distributed, we report both the mean and median of the distribution in the latter analyses.

Of the 4,549 participants in the SHS cohort, 2,127 (46.8%) were diabetic at baseline. Of these participants, 12 with a baseline LEA caused by something other than diabetes (trauma [n = 5], other [n = 1], and unknown [n = 6]) and 7 participants with incomplete or conflicting LEA or diabetes duration data were excluded from the analysis. The analysis sample therefore consists of 2,108 individuals with diabetes.

Of the 2,108 diabetic participants, 134 (6.4%) had an LEA at baseline. Of these, 21 (15.7%) had bilateral LEAs. Table 1 compares baseline characteristics of diabetic SHS participants with an LEA to those without. Diabetic participants with an LEA had longer duration of diabetes and higher systolic blood pressure, HbA_1c, and fibrinogen than those without, as well as more albuminuria and a higher prevalence of both high and low ABI. Amputations were substantially more frequent among men and among those in Arizona (7.2%) than among those in the Dakotas (0.95%) or Oklahoma (0.88%). Diabetic participants with LEA had lower BMI and LDL cholesterol than those without. No relationship was observed between LEA and either smoking or at-risk drinking.

Diabetic women with a lower-extremity amputation were older (60.3 vs. 56.1 years) and had longer duration of diabetes (22.9 vs. 16.9) compared with men with an LEA, but they reported smoking less frequently than men (58 vs. 79%, P < 0.01 for all) (data not shown).

The 134 participants with a lower-extremity amputation were categorized according to the most proximal amputation location (22 participants had bilateral LEAs): 1 was an AKA, 53 were BKAs, and 80 were of toes. Figure 1 illustrates the distribution of baseline LEA location and subsequent disposition of these participants throughout follow-up. Of the participants with LEA, 23% experienced a recurrent LEA during follow-up, and these consisted of a mix of ipsilateral and contralateral procedures. With the exception of age, no differences in risk factors were observed between participants who experienced a recurrent LEA and those who did not (data not shown). However, 38% of participants with a baseline LEA died before the second examination, and this rose to 76% at study end. The high mortality rate among individuals with an LEA precluded meaningful multivariable analysis of risk factors for recurrent lower-extremity amputations.

Table 2 shows crude incidence rates and HRs for all-cause and CVD mortality by LEA and sex. Diabetic women with an LEA had a 4.8-fold increased risk of all-cause and 6.7-fold increased risk of CVD mortality compared with diabetic women without an LEA. Diabetic men with a lower-extremity amputation also had significantly increased risk of all-cause and CVD mortality, but the magnitude of the associations was lower than that for women (3.0 and 2.1, respectively).

Cox proportional hazards models of all-cause mortality are based on data for 1,984 participants with complete data for all risk factors, and CVD mortality models are based on data for 1,963 participants. Of the 124 individuals excluded from all-cause mortality analyses, 35 had a baseline LEA, and of the 145 individuals excluded from multivariable analyses of CVD mortality, 29 had a baseline LEA. Most of the missing observations were from individuals with missing data for ABI or albuminuria.

In the whole study sample, lower-extremity amputation persisted as a significant predictor of both all-cause mortality (HR 2.2, 95% CI 1.7–2.9) and CVD mortality (1.9, 1.2–2.9) after adjustment for risk factors that differed between groups in initial analyses. We then identified the most parsimonious models to describe the effect of LEA on mortality in men and women separately, and these risk estimates are illustrated in Fig. 2. Women with a lower-extremity amputation had a 2.9 (CI 2.1–4.2) and 3.8 (2.3–6.3) times greater risk of all-cause and CVD mortality, respectively, compared with women without. LEA was a significant predictor of all-cause mortality in men (HR 1.7, CI 1.2–2.5) but was unrelated to CVD mortality (1.1, 0.5–2.2) in men. We also constructed sex-specific models using the same covariates for men and women (rather than the most parsimonious model) and observed no meaningful difference in the LEA effect on mortality (all-cause mortality: women 2.6, men 1.8; CVD mortality: women 3.2, men 1.0).

Sex-specific models suggested a potential interaction between lower-extremity amputation and sex on mortality risk. Formal testing of this interaction indicated no sex effect on the association between LEA and all-cause mortality. However, we observed a significant interaction (P = 0.02) between LEA and sex on CVD mortality risk that indicated higher LEA-associated CVD mortality risk among women with a lower-extremity amputation compared with men with an LEA. The HR for CVD mortality in women with an LEA was 3.0 (CI 1.7–5.2), whereas the HR for men with an LEA was 1.1 (0.6–2.2).

A trend of increasing mortality with more proximal LEA was evident. All-cause mortality rates were 34.6, 114.8, and 143.7 per 1,000 person-years for diabetic people without an LEA and among those with toe and below-the-knee amputation, respectively. A similar pattern was evident for CVD mortality.

Figure 3 shows survival distribution functions over 8.7 ± 2.9 years of follow-up, according to baseline LEA site. The curves show higher mortality among diabetic participants with an LEA compared with those without, and they suggest an association between proximity of LEA and increased mortality. However, survival among participants with a toe amputation did not significantly differ from those with a BKA (P = 0.27). Survival of diabetic individuals without an LEA was significantly better (P < 0.01) than that of each of the LEA groups, and survival was most favorable among individuals without diabetes.

Diabetic participants with a toe amputation had a 2.0 (1.4–2.7) and 1.5 (0.9–2.6) times greater risk of all-cause and CVD mortality, respectively, compared with those without a lower-extremity amputation. The HR was greater in those with a BKA for both all-cause (2.3, 1.5–3.4) and CVD (2.9, 1.6–5.6) mortality. Consistent with these data, there were substantial differences in diabetes duration according to presence and site of amputation, with mean diabetes duration of 11.9, 18.6, and 21.1 years for participants with diabetes and no lower-extremity amputation, participants with toe amputation, and those with a BKA, respectively.

Among 2,108 diabetic participants in the SHS baseline examination, 134 (6.4%) had a prevalent nontraumatic amputation that was attributable to diabetes. The prevalence of LEA in this report exceeds that in other studies of American Indians. A study of Oklahoma Indians (10) and another of Plains Indians (26) reported an LEA prevalence of 2.1 and 3.6% among diabetic and nondiabetic participants combined, and a different study reported a 4.0% prevalence of LEA among diabetic Indians in the Pacific Northwest (7). A probable reason for the differences in LEA prevalence between diabetic participants in our study and previous ones is the older age and the longer mean duration of diabetes in the SHS population (11.9 ± 9.0 years) compared with the populations of other studies in which these data were available (6.9 ± 6.4 years) (10). One study reported a 10.3% prevalence of LEA among diabetic Tohono O’odham Indians in Arizona (8), a figure very close to the center-specific prevalence of lower-extremity amputation for the diabetic individuals in the Arizona Strong Heart Center (10.6%).

During an average of 8.7 ± 2.9 years of follow-up, there was 76% mortality among individuals with a baseline lower-extremity amputation. Diabetic participants with an LEA had all-cause mortality rates that were 7.2 times higher than in all participants without and 3.6 times higher than diabetic participants without. These results are consistent with previous studies showing poor prognosis among individuals with an LEA. The World Health Organization (WHO) Multinational Study of Vascular Disease in Diabetes showed a mortality rate of 22.7 per 1,000 person-years among American Indians with type 2 diabetes (6). This rate is considerably lower than our mortality rate of 129 per 1,000 person-years. The WHO mortality rates, however, were based on categorization of the exposure that included gangrene or amputation. This is less specific than our exposure, which included only documented amputation attributable to diabetes. Thus, it is not surprising that mortality risk estimates from this report are higher than those reported by the WHO Multinational Study. It would be interesting to examine the effect on mortality of vascular reconstruction before or after a lower-extremity amputation. Unfortunately, these data are not available in the SHS, making the effect of these procedures on our risk estimates impossible to quantify with certainty.

Although multivariable models attenuated the crude LEA effect on mortality, LEA remained a significant predictor of all-cause mortality in both men and women and of CVD mortality among women. These results suggest that the effect of LEA on mortality was not entirely attributable to diabetes-related comorbidities, diabetes duration, obesity, or differences among the study centers in this cohort.

A novel finding in this report is the significant interaction between lower-extremity amputation and female sex on CVD mortality risk. In sex-stratified analyses, the CVD HR for women was 3.8 (2.3–6.3), but it was 1.2 (0.6–2.3) for men. Formal testing of the lower-extremity amputation × sex interaction yielded significant findings suggesting that LEA confers more risk of CVD mortality in women relative to men. The HR for CVD mortality in women with an LEA was 3.0 (CI 1.7–5.2), whereas the HR for men with an LEA was 1.1 (CI 0.6–2.2). It is important to note that the LEA × sex interaction on CVD mortality occurs against a backdrop of a twofold increased risk of CVD mortality in men in this cohort. Thus, it is possible that lower-extremity amputation distinguishes a group of diabetic women at particularly high risk of CVD, whereas it does not add significantly to the already elevated CVD mortality risk in men.

Although smoking is a known risk factor for LEA and mortality, our analyses did not indicate that ever-smokers were at increased risk of either all-cause or CVD mortality. A likely explanation for this finding is that ever-smokers in the SHS cohort smoke ∼11 cigarettes per day compared with 20 per day nationally (27). Tobacco use also differs across the SHS centers and by sex, with daily cigarette use among Arizona women as low as six per day. Thus, the effects of smoking on mortality may be less in American Indians because of the lower dose of tobacco. Although there is no evidence supporting the absence of negative health effects of tobacco use in this cohort, these effects may be overwhelmed by a combination of the strength of other risk factors and the relatively low dose of tobacco among ever-smokers.

LEA is a risk factor for mortality probably because it is a highly specific marker of damage to large and small vessels, as well as peripheral nerve damage resulting from long-standing diabetes. This hypothesis is supported by our data, which show longer duration of diabetes among individuals with an LEA, a high prevalence of peripheral arterial disease and renal dysfunction among individuals with a lower-extremity amputation, a relation of these factors with mortality in our cohort, and the persistent effect of LEA on mortality after adjustment for these factors.

LEA may also be a marker for the effect of nonbiological factors on health outcomes. Individuals with an LEA may receive less adequate health care or have poor diabetes management skills compared with those without one; both factors would increase mortality risk. The majority of participants in the SHS receive health care from the IHS, which has standards for both clinical care and diabetes education. However, standards of care for diabetes are often difficult to meet in rural areas served by the IHS, as they are in the rest of rural America. An additional point is that the site of amputation may be related to practice patterns or other professional differences in how ischemic vessels are handled.

The greater prevalence of lower-extremity amputation in the Arizona center, along with greater prevalence of insensitivity to the monofilament from previous work in SHS (28), suggest that environmental or other differences between centers or differences in genetic susceptibility to neuropathy may lead to elevated risk of LEA in Pima Indians compared with other American Indian tribes. A likely explanation for the difference in both insensitivity to the monofilament and the increased prevalence of LEA is the longer mean duration of diabetes in the Arizona center (14.8 years) compared with the Oklahoma (10.7 years) and Dakota (9.8 years) centers. However, adjustment for center had no effect on the strength of association between LEA and mortality.

Foot ulceration and LEA in diabetic individuals occur most often as a result of the combined effects of vascular and nerve damage resulting from exposure to a prolonged hyperglycemic environment, pathways that are consistent with data presented in this report (29,30). Monofilament testing measures large nerve fiber function, an established risk factor for LEA (31,32). Although no data on neuropathy were collected at the baseline SHS exam, monofilament testing was conducted at the second SHS examination, which occurred ∼5 years after baseline. These data showed that duration of diabetes was significantly associated with insensitivity to the monofilament (28), a finding that is consistent with data from the current analysis showing an association between LEA and duration of diabetes.

We are unaware of previous studies that have quantified the strength of association between the anatomical site of LEA and subsequent mortality risk. The adjusted HRs for all-cause mortality for toe and BKA compared with those without a lower-extremity amputation were 2.0 and 2.3, respectively. Although survival analysis showed significant differences in survival between diabetic subjects with and without an LEA, the graded survivorship among lower-extremity amputation sites did not reach statistical significance. The lack of significant findings was due to limited statistical power associated with the small number of site-specific LEAs.

Participants who reported diabetes at baseline had an average diabetes duration of 12 years. Diabetes duration was longer in those with an LEA, and it increased as the site of amputation became more proximal. Participants with a toe amputation had an average diabetes duration of 19 years; participants with a BKA had an average diabetes duration of 21 years, and the participant with an AKA had diabetes for 34 years. These findings highlight the effect of duration of diabetes on severity of diabetic complications and on mortality risk, and they are consistent with the higher prevalence of LEA in Arizona, the center with the longest mean duration of diabetes.

Increases in the occurrence of type 2 diabetes in American Indian children may “push back the clock” with respect to development of complications. Diabetic complications such as LEA are likely to begin appearing earlier in life because of the earlier onset of diabetes among today’s youth. Without intensive prevention efforts, today’s children will spend a greater proportion of their adult lives coping with the social and physical difficulties associated with an LEA and other diabetes complications. Our findings of excessive mortality associated with amputation suggest that future research should focus on efficacy of interventions such as surveillance of foot health and patient self-care and education.

This study was supported by grants U01-HL-41642, U01-HL-41652, and U01-HL-41654 from the National Heart, Lung, and Blood Institute.

The authors acknowledge the assistance and cooperation of the AkChin Tohono O’odham (Papago)/Pima, Apache, Caddo, Cheyenne River Sioux, Comanche, Delaware, Spirit Lake, Fort Sill Apache, Gila River Pima/Maricopa, Kiowa, Oglala Sioux, Salt River Pima/Maricopa, and Wichita Indian communities, without whose support this study would not have been possible. The authors also thank the IHS hospitals and clinics at each center and Betty Jarvis, RN, Tauqeer Ali, MD, PhD, and Marcia O’Leary, RN, directors of the SHS clinics and their staffs. Dr. Chris Burd provided support for the study of foot disease in the SHS.