Epidemiology/Health Services/Psychosocial Research ORIGINAL ARTICLE

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

1 3 HELAINE E. RESNICK, PHD, MPH FREDERICK K. NESS, MD merican Indians have very high 1 4 ELIZABETH A. CARTER, MPH THOMAS K. WELTY, MD rates of type 2 , and rates of 1 5 ROBERT LINDSAY, MB, PHD ELISA T. LEE, PHD 2 1 diabetes-associated morbidity and USAN ENLY PHD, RN ARBARA OWARD PHD A S J. H , B V. H , mortality consistently exceed those of other ethnic groups (1–4). One of the most disabling complications of diabetes is lower-extremity amputation (LEA). The LEA event and its aftermath are char- OBJECTIVE — To compare risk of all-cause and (CVD) mortality in acterized by pain, immobility, and people with a lower-extremity amputation (LEA) attributable to diabetes and people without an LEA. changes in body image and social func- tioning. Work restrictions are common, RESEARCH DESIGN AND METHODS — The Strong Heart Study is a study of CVD and most individuals with LEA require and its risk factors in 13 American-Indian communities. LEA was ascertained at baseline by extensive life-long health services. Lower- direct examination of the legs and feet. Mortality surveillance is complete through 2000. extremity amputations occur with signif- icantly greater frequency in diabetic RESULTS — Of 2,108 participants with diabetes at baseline, 134 participants (6.4%) had an American Indians than in diabetic people LEA. Abnormal ankle-brachial index (53%), albuminuria (87%), and long diabetes duration of other ethnicities (5,6). (mean 19.8 years) were common among diabetic subjects with LEA. Mean diabetes duration A number of studies have demon- among diabetic participants without LEA and in those with toe and below-the-knee amputations was 11.9, 18.6, and 21.1 years, respectively. During 8.7 (Ϯ2.9) years of follow-up, 102 of the strated the high of lower- participants with LEA (76%) died from all causes and 35 (26%) died from CVD. Of the 1,974 extremity amputations in individual diabetic participants without LEA at baseline, 604 (31%) died from all causes and 206 (10%) American-Indian tribes (7–9), and several died from CVD. The unadjusted hazard ratios (HRs) for all-cause and CVD mortality in diabetic early studies reported prospective data on participants with LEA compared with those without were 4.0 and 4.1, respectively. Adjusting for high LEA rates among American Indians known and suspected confounders, LEA persisted as a predictor of all-cause (HR 2.2, 95% CI over time (10–13). In studies where risk 1.7–2.9) and CVD mortality (HR 1.9, 95% CI 1.3–2.9). We observed a significant interaction factor information was available, increas- between baseline LEA and sex on CVD mortality, with female sex conferring added risk of CVD ing age, glucose levels (6,10,11,14), mortality. longer duration of diabetes (10,11), male CONCLUSIONS — LEA is a potent predictor of all-cause and CVD mortality in diabetic sex (7,10,11,14), and presence of reti- American Indians. The combination of female sex and LEA is associated with greater risk of CVD nopathy (6,10,11,14) have been shown to mortality than either factor alone. increase lower-extremity amputation risk in American Indians. Data on LEA pat- Diabetes Care 27:1286–1293, 2004 terns over time, risk factors associated with recurrent LEAs, and the relation of lower-extremity amputation to mortality in American Indians are sparse. ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● This study was designed to define in a From the 1MedStar Research Institute, Hyattsville, Maryland; the 2School of Nursing, University of Minne- population-based study of Indian people 3 sota, Minneapolis, Minnesota; Mille Lacs Band of Ojibwe Indians, Diabetes Program, Mille Lacs, Minnesota; the prevalence of LEA in diabetic patients, 4Missouri Breaks Research, Timber Lake, South Dakota; and the 5University of Health Sciences Center, Oklahoma City, Oklahoma. examine risk factors associated with prev- Address correspondence and reprint requests to Helaine E. Resnick, PhD, MPH, Department of Epide- alent and recurrent LEAs, define rates of miology and Statistics, MedStar Research Institute, 6495 New Hampshire Ave., Suite 201, Hyattsville, MD all-cause and cardiovascular disease 20783. E-mail: [email protected]. (CVD) mortality among individuals with Received for publication 5 February 2004 and accepted in revised form 12 March 2004. Abbreviations: ABI, ankle brachial index; AKA, above-the-knee amputation; BKA, below-the-knee am- an LEA, and study the relationship of low- putation; CVD, cardiovascular disease; IHS, Indian Health Service; LEA, lower-extremity amputation; SHS, er-extremity amputation location to mor- Strong Heart Study; WHO, World Health Organization. tality. We hypothesized that among The opinions expressed in this paper are those of the authors and do not necessarily reflect the view of the diabetic individuals, diabetes duration, IHS. A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion male sex, and smoking would increase the factors for many substances. odds of prevalent LEAs and the risk of © 2004 by the American Diabetes Association. recurrent LEAs, that LEA would be asso-

1286 DIABETES CARE, VOLUME 27, NUMBER 6, JUNE 2004 Resnick and Associates ciated with substantially increased mor- (5,6,9,10,14,18). Participants were con- and ICD-9 codes were applied centrally tality risk, and that proximal LEAs would sidered hypertensive if they had a systolic by a nosologist (25). The SHS Mortality be more strongly associated with mortal- blood pressure Ն140 mmHg, a diastolic Review Committee independently con- ity than distal LEAs. blood pressure Ն90 mmHg, or if they firmed the cause of each study participant were taking antihypertension medication. death by review of autopsy reports, med- RESEARCH DESIGN AND Smoking history was determined by ical records, and informant interviews, METHODS — The Strong Heart questionnaire. Excessive alcohol con- using published criteria to ascertain Study (SHS) was initiated in 1988 to in- sumption can cause or aggravate neurop- deaths due to fatal , vestigate CVD and its risk factors in Amer- athy. In this report, at-risk drinking was coronary heart disease, and (25). ican Indians (15). The design, methods, defined as a report of either Ն5 drinks on and laboratory techniques of the SHS one occasion in the past month or Ն14 Statistical methods have been reported previously (15–17). drinks in a typical week. We used ␹2 and Student’s t tests to exam- The SHS cohort consists of 4,549 partici- ine differences in proportions and means pants aged 45–74 from , Okla- Risk factors for all-cause and CVD between categorical and continuous vari- homa, and South and North Dakota who mortality ables, respectively. Initial comparisons were seen at the baseline examination, Methods for measurement of CVD risk focused on differences in risk factors be- which was conducted between 1989 and factors, including BMI, waist circumfer- tween participants with and without an 1992. The second and third examinations ence, ankle and arm blood pressures, and LEA at the baseline examination. Subse- were conducted between 1993 and 1995 renal function have been described and quent analyses focused on differences in and between 1997 and 1999, respec- laboratory methods have been published risk factors between participants with an tively. Nonparticipants were similar to previously (15,16,19). , tri- LEA who did and did not experience a participants in age, BMI, and self- glycerides, and fasting glucose were de- recurrent amputation. reported frequency of diabetes and hyper- termined by enzymatic methods using a Person-years were calculated from tension (17). The SHS was approved by Hitachi chemistry analyzer and consis- the date of the baseline examination to the the institutional review boards of partici- tent, standardized reagents (Boehringer date of a fatal event, the date of the last pating institutions, and all participants Mannheim Diagnostics, Indianapolis, exam if the participant was lost to follow provided informed consent. Surveillance IN). Fasting insulin and fibrinogen were up (n ϭ 7; 0.2%), or 31 December 2000. for cause-specific mortality is complete measured by established methods Sex-specific rates of all-cause and CVD through 31 December 2000. (20,21). Urinary albumin excretion was mortality were calculated according to estimated by the ratio of albumin (mg) to baseline LEA status, and crude rate ratios Evaluation of LEA creatinine (g) in a morning urine sample. quantifying the effect of LEA on mortality At each of the three SHS examinations, Microalbuminuria was defined as a ratio were estimated. Cox regression models trained examiners collected amputation of urinary albumin (mg/ml) to creatinine were used to determine whether LEA pre- data from direct observation of both legs. (g/ml) of 30–299 mg/g, and macroalbu- dicted all-cause or CVD mortality, adjust- Data from this exam included presence/ minuria was defined as a ratio of Ͼ300 ing for potential confounders. These absence and cause of missing extremities. mg/g. analyses are presented for the whole sam- The location of missing extremities was An ankle brachial index (ABI) of ple and stratified by sex. We tested the coded by study personnel as toe, below Ͻ0.90 (“low ABI”) is 95% sensitive and proportional hazards assumption by plot- the knee (BKA), or above the knee (AKA). 99% specific for angiographically docu- ting hazard curves for LEA against the log Participants were asked to attribute each mented peripheral arterial disease of time. No evidence was found that this missing extremity to a specific factor. (22,23), and it has been used in previous assumption was violated. These attributions were coded as diabe- work in the SHS (19). Although it is not a Survival analysis was used to visualize tes, trauma, congenital, other, and un- definitive assessment of medial arterial survival patterns among diabetic partici- known. In this report, LEA refers to calcification, data from the SHS (24) show pants with an LEA at specific sites (toe, amputations that were caused by diabe- that ABI Ͼ1.40 (“high ABI”) is associated BKA) compared with those without an tes. with a similar level of CVD risk as ABI LEA and individuals without diabetes. Ͻ0.90. Baseline ABI was used to catego- The log-rank test was used to evaluate dif- Definition of diabetes rize three groups of individuals: those ferences in survivorship among these cat- Participants were categorized as diabetic with “low” ABI (Ͻ0.90), those in the “nor- egories, and the Cox model was used to if they had a fasting glucose Ն126 mg/dl, mal” range of ABI (0.90 Յ ABI Յ 1.40), quantify adjusted hazard ratios (HRs) as- reported use of hypoglycemic medica- and those with “high” ABI (Ͼ1.40). sociated with LEA at different anatomical tions, or reported diabetes at the baseline sites. We examined duration of diabetes examination. Ascertainment of all-cause and CVD in relation to the location of LEAs. Be- mortality cause diabetes duration was not normally Measurement of baseline risk factors Deaths among cohort members were distributed, we report both the mean and Previous studies of LEA in American In- identified through tribal and Indian median of the distribution in the latter dians and other populations provide a ra- Health Service (IHS) hospital records and analyses. tionale for selection of potential LEA risk through direct contact with participant factors. These include age, glycemic con- families (15,17). Death certificates were RESULTS — Of the 4,549 participants trol, smoking, , and male sex obtained from state health departments, in the SHS cohort, 2,127 (46.8%) were

DIABETES CARE, VOLUME 27, NUMBER 6, JUNE 2004 1287 Amputation and mortality in American Indians

2,108 ؍ Table 1—Characteristics of SHS participants with diabetes, by baseline LEA status, n

Difference in means or crude Characteristic Diabetic subjects with LEA Diabetic, no LEA OR (95% CI) P n 134 1,974 Age (years) 58.1 57.0 1.1 (Ϫ0.3 to 2.5) 0.13 Sex (male) 70 (52) 712 (36) 1.9 (1.4–2.8) Ͻ0.01 Center North/South Dakota 14 (10) 514 (26) 1.0 — Oklahoma 13 (10) 560 (28) 0.9 (0.4–1.8) 0.68 Arizona 107 (80) 900 (46) 4.4 (2.5–7.7) Ͻ0.01 Duration of diabetes (years)* 19.8 11.9 7.9 (6.2–9.5) Ͻ0.01 Undiagnosed diabetes†—480 (24) N/A —

HbA1c (%) 9.0 8.5 0.5 (0.1–1.0) 0.02 Hypertension‡ 69 (52) 893 (45) 1.3 (0.9–1.8) 0.16 Systolic blood pressure (mmHg) 138 131 6.5 (2.9–10.2) Ͻ0.01 Diastolic blood pressure 78 77 0.7 (Ϫ1.0 to 2.5) 0.42 (mmHg) Total cholesterol (mg/dl) 187 190 Ϫ2.7 (Ϫ11.0 to 5.5) 0.56 Triglycerides (mg/dl) 198 177 20.1 (Ϫ14.9 to 55.0) 0.19 HDL cholesterol (mg/dl) 44 43 1.5 (Ϫ0.7 to 3.6) 0.34 LDL cholesterol (mg/dl) 101 110 Ϫ9.9 (Ϫ16.1 to Ϫ3.8) Ͻ0.01 Fibrinogen ln (mg/dl) 5.87 5.74 0.14 (0.09–0.18) Ͻ0.01 BMI (kg/m2) 29.3 32.2 Ϫ2.9 (Ϫ4.1 to Ϫ1.7) Ͻ0.01 At-risk drinking§ 36 (27) 419 (21) 1.4 (0.90–2.0) 0.13 Ever smoker 92 (69) 1,323 (67) 1.1 (0.7–1.6) 0.73 Albuminuria Normal 16 (13) 971 (51) 1.0 — Microalbuminuria 26 (22) 589 (31) 2.7 (1.4–5.0) Ͻ0.01 Macroalbuminuria 77 (65) 344 (18) 13.6 (7.8–23.6) Ͻ0.01 ABI group¶ Ͻ0.90 (low) 11 (9) 120 (6) 2.7 (1.4–5.2) Ͻ0.01 Ն0.90 to Յ1.40 (normal) 55 (47) 1,600 (82) 1.0 — Ͼ1.40 (high) 52 (44) 230 (12) 6.6 (4.4–9.9) Ͻ0.01 Data are n (%), unless otherwise indicated. *Data available for 126 diabetic subjects with LEA and 1,423 participants with reported diabetes and no LEA. †No reported diabetes and fasting glucose Ն126 mg/dl. ‡Systolic blood pressure Ն140 mmHg or diastolic blood pressure Ն90 mmHg or taking antihypertension medication. §Report of 5 or more drinks on one occasion or 14 or more drinks in a single week. Data available for 119 diabetic subjects with LEA and 1,904 diabetic subjects with no LEA. ¶Data available for 118 diabetic subjects with LEA and 1,950 diabetic subjects with no LEA. diabetic at baseline. Of these participants, ABI. Amputations were substantially eral LEAs): 1 was an AKA, 53 were BKAs, 12 with a baseline LEA caused by some- more frequent among men and among and 80 were of toes. Figure 1 illustrates thing other than diabetes (trauma [n ϭ 5], those in Arizona (7.2%) than among the distribution of baseline LEA location other [n ϭ 1], and unknown [n ϭ 6]) and those in the Dakotas (0.95%) or Okla- and subsequent disposition of these par- 7 participants with incomplete or con- homa (0.88%). Diabetic participants with ticipants throughout follow-up. Of the flicting LEA or diabetes duration data LEA had lower BMI and LDL cholesterol participants with LEA, 23% experienced a were excluded from the analysis. The than those without. No relationship was recurrent LEA during follow-up, and analysis sample therefore consists of observed between LEA and either smok- these consisted of a mix of ipsilateral and 2,108 individuals with diabetes. ing or at-risk drinking. contralateral procedures. With the excep- Of the 2,108 diabetic participants, Diabetic women with a lower- tion of age, no differences in risk factors 134 (6.4%) had an LEA at baseline. Of extremity amputation were older (60.3 were observed between participants who these, 21 (15.7%) had bilateral LEAs. Ta- vs. 56.1 years) and had longer duration of experienced a recurrent LEA and those ble 1 compares baseline characteristics of diabetes (22.9 vs. 16.9) compared with who did not (data not shown). However, diabetic SHS participants with an LEA to men with an LEA, but they reported 38% of participants with a baseline LEA those without. Diabetic participants with smoking less frequently than men (58 vs. died before the second examination, and an LEA had longer duration of diabetes 79%, P Ͻ 0.01 for all) (data not shown). this rose to 76% at study end. The high and higher systolic blood pressure, The 134 participants with a lower- among individuals with an HbA1c, and fibrinogen than those with- extremity amputation were categorized LEA precluded meaningful multivariable out, as well as more albuminuria and a according to the most proximal amputa- analysis of risk factors for recurrent low- higher prevalence of both high and low tion location (22 participants had bilat- er-extremity amputations.

1288 DIABETES CARE, VOLUME 27, NUMBER 6, JUNE 2004 Resnick and Associates

Figure 1—Most proximal site of LEA at baseline and subsequent disposition of SHS diabetic patients with an LEA, 1988–2000, n ϭ 134.

Table 2 shows crude rates all risk factors, and CVD mortality models ment for risk factors that differed between and HRs for all-cause and CVD mortality are based on data for 1,963 participants. groups in initial analyses. We then iden- by LEA and sex. Diabetic women with an Of the 124 individuals excluded from all- tified the most parsimonious models to LEA had a 4.8-fold increased risk of all- cause mortality analyses, 35 had a base- describe the effect of LEA on mortality in cause and 6.7-fold increased risk of CVD line LEA, and of the 145 individuals men and women separately, and these mortality compared with diabetic women excluded from multivariable analyses of risk estimates are illustrated in Fig. 2. without an LEA. Diabetic men with a low- CVD mortality, 29 had a baseline LEA. Women with a lower-extremity amputa- er-extremity amputation also had signifi- Most of the missing observations were tion had a 2.9 (CI 2.1–4.2) and 3.8 (2.3– cantly increased risk of all-cause and CVD from individuals with missing data for 6.3) times greater risk of all-cause and mortality, but the magnitude of the asso- ABI or albuminuria. CVD mortality, respectively, compared ciations was lower than that for women In the whole study sample, lower- with women without. LEA was a signifi- (3.0 and 2.1, respectively). extremity amputation persisted as a sig- cant predictor of all-cause mortality in Cox proportional hazards models of nificant predictor of both all-cause men (HR 1.7, CI 1.2–2.5) but was unre- all-cause mortality are based on data for mortality (HR 2.2, 95% CI 1.7–2.9) and lated to CVD mortality (1.1, 0.5–2.2) in 1,984 participants with complete data for CVD mortality (1.9, 1.2–2.9) after adjust- men. We also constructed sex-specific

Table 2—Person-years at risk, number of deaths, and crude incidence of all-cause and CVD mortality among American Indians with diabetes, 2,108 ؍ by baseline LEA status, n

Incidence Incidence of of death/ CVD death/ 1,000 Number of 1,000 Number of Person-years person- All-cause CVD person- CVD hazard n deaths at risk years hazard ratio deaths years ratio Women Prevalent LEA 64 50 379 131.9 4.8 (3.6–6.5) 22 58.1 6.7 (4.2–10.6) No LEA 1,262 349 11,431 30.5 1.0 115 10.1 1.0 Total 1,326 399 11,810 33.8 — 137 11.6 — Men Prevalent LEA 70 52 428 121.5 3.0 (2.3–4.1) 13 30.4 2.1 (1.2–3.8) No LEA 712 255 6,023 42.3 1.0 91 15.1 1.0 Total 782 307 6,451 47.6 — 104 16.1 — Men and women Prevalent LEA 134 102 807 126.4 4.0 (3.2–4.9) 35 43.4 4.1 (2.8–5.8) No LEA 1,974 604 17,454 34.6 1.0 206 11.8 1.0 Total 2,108 706 18,261 38.7 — 241 13.2 —

DIABETES CARE, VOLUME 27, NUMBER 6, JUNE 2004 1289 Amputation and mortality in American Indians

and increased mortality. However, sur- vival among participants with a toe am- putation did not significantly differ from those with a BKA (P ϭ 0.27). Survival of diabetic individuals without an LEA was significantly better (P Ͻ 0.01) than that of each of the LEA groups, and survival was most favorable among individuals with- out diabetes. Diabetic participants with a toe am- Figure 2—Risk of all-cause and CVD mortality in diabetic American Indians with a lower- putation had a 2.0 (1.4–2.7) and 1.5 extremity amputation compared with those without. All-cause mortality model: *adjusted for age, (0.9–2.6) times greater risk of all-cause BMI, albuminuria, LDL cholesterol, center, and diabetes duration; adjusted for age, fibrinogen, and CVD mortality, respectively, com- high ABI, low ABI, albuminuria, LDL cholesterol, and diabetes duration. CVD mortality model: pared with those without a lower- *adjusted for age, albuminuria, triglycerides, LDL cholesterol, and diabetes duration; adjusted extremity amputation. The HR was for age, fibrinogen, high ABI, low ABI, albuminuria, LDL cholesterol, center, and diabetes dura- greater in those with a BKA for both all- tion. cause (2.3, 1.5–3.4) and CVD (2.9, 1.6– 5.6) mortality. Consistent with these data, models using the same covariates for men men with an LEA. The HR for CVD mor- there were substantial differences in dia- and women (rather than the most parsi- tality in women with an LEA was 3.0 (CI betes duration according to presence and monious model) and observed no mean- 1.7–5.2), whereas the HR for men with an site of amputation, with mean diabetes ingful difference in the LEA effect on LEA was 1.1 (0.6–2.2). duration of 11.9, 18.6, and 21.1 years for mortality (all-cause mortality: women A trend of increasing mortality with participants with diabetes and no lower- 2.6, men 1.8; CVD mortality: women 3.2, more proximal LEA was evident. All- extremity amputation, participants with men 1.0). cause mortality rates were 34.6, 114.8, toe amputation, and those with a BKA, Sex-specific models suggested a po- and 143.7 per 1,000 person-years for di- respectively. tential interaction between lower- abetic people without an LEA and among extremity amputation and sex on those with toe and below-the-knee ampu- CONCLUSIONS — Among 2,108 di- mortality risk. Formal testing of this inter- tation, respectively. A similar pattern was abetic participants in the SHS baseline ex- action indicated no sex effect on the asso- evident for CVD mortality. amination, 134 (6.4%) had a prevalent ciation between LEA and all-cause Figure 3 shows survival distribution nontraumatic amputation that was attrib- mortality. However, we observed a signif- functions over 8.7 Ϯ 2.9 years of follow- utable to diabetes. The prevalence of LEA icant interaction (P ϭ 0.02) between LEA up, according to baseline LEA site. The in this report exceeds that in other studies and sex on CVD mortality risk that indi- curves show higher mortality among dia- of American Indians. A study of Okla- cated higher LEA-associated CVD mortal- betic participants with an LEA compared homa Indians (10) and another of Plains ity risk among women with a lower- with those without, and they suggest an Indians (26) reported an LEA prevalence extremity amputation compared with association between proximity of LEA of 2.1 and 3.6% among diabetic and non-

Figure 3—Survival distribution function by baseline diabetes and LEA status, n ϭ 4,422

1290 DIABETES CARE, VOLUME 27, NUMBER 6, JUNE 2004 Resnick and Associates diabetic participants combined, and a dif- bidities, diabetes duration, obesity, or dif- and the persistent effect of LEA on mor- ferent study reported a 4.0% prevalence ferences among the study centers in this tality after adjustment for these factors. of LEA among diabetic Indians in the Pa- cohort. LEA may also be a marker for the ef- cific Northwest (7). A probable reason for A novel finding in this report is the fect of nonbiological factors on health the differences in LEA prevalence be- significant interaction between lower- outcomes. Individuals with an LEA may tween diabetic participants in our study extremity amputation and female sex on receive less adequate health care or have and previous ones is the older age and the CVD mortality risk. In sex-stratified anal- poor diabetes management skills com- longer mean duration of diabetes in the yses, the CVD HR for women was 3.8 pared with those without one; both fac- SHS population (11.9 Ϯ 9.0 years) com- (2.3–6.3), but it was 1.2 (0.6–2.3) for tors would increase mortality risk. The pared with the populations of other stud- men. Formal testing of the lower- majority of participants in the SHS receive ies in which these data were available extremity amputation ϫ sex interaction health care from the IHS, which has stan- (6.9 Ϯ 6.4 years) (10). One study re- yielded significant findings suggesting dards for both clinical care and diabetes ported a 10.3% prevalence of LEA among that LEA confers more risk of CVD mor- education. However, standards of care for diabetic Tohono O’odham Indians in Ar- tality in women relative to men. The HR diabetes are often difficult to meet in rural izona (8), a figure very close to the center- for CVD mortality in women with an LEA areas served by the IHS, as they are in the specific prevalence of lower-extremity was 3.0 (CI 1.7–5.2), whereas the HR for rest of rural America. An additional point amputation for the diabetic individuals in men with an LEA was 1.1 (CI 0.6–2.2). It is that the site of amputation may be re- ϫ the Arizona Strong Heart Center (10.6%). is important to note that the LEA sex lated to practice patterns or other profes- Ϯ During an average of 8.7 2.9 years interaction on CVD mortality occurs sional differences in how ischemic vessels of follow-up, there was 76% mortality against a backdrop of a twofold increased are handled. among individuals with a baseline lower- risk of CVD mortality in men in this co- The greater prevalence of lower- extremity amputation. Diabetic partici- hort. Thus, it is possible that lower- extremity amputation in the Arizona cen- pants with an LEA had all-cause mortality extremity amputation distinguishes a ter, along with greater prevalence of rates that were 7.2 times higher than in all group of diabetic women at particularly insensitivity to the monofilament from participants without and 3.6 times higher high risk of CVD, whereas it does not add previous work in SHS (28), suggest that than diabetic participants without. These significantly to the already elevated CVD environmental or other differences be- results are consistent with previous stud- mortality risk in men. tween centers or differences in genetic ies showing poor prognosis among indi- Although smoking is a known risk susceptibility to neuropathy may lead to viduals with an LEA. The World Health factor for LEA and mortality, our analyses elevated risk of LEA in Pima Indians com- Organization (WHO) Multinational did not indicate that ever-smokers were at pared with other American Indian tribes. Study of Vascular Disease in Diabetes increased risk of either all-cause or CVD A likely explanation for the difference in showed a mortality rate of 22.7 per 1,000 mortality. A likely explanation for this both insensitivity to the monofilament person-years among American Indians finding is that ever-smokers in the SHS with (6). This rate is con- cohort smoke ϳ11 cigarettes per day and the increased prevalence of LEA is the siderably lower than our mortality rate of compared with 20 per day nationally longer mean duration of diabetes in the 129 per 1,000 person-years. The WHO (27). Tobacco use also differs across the Arizona center (14.8 years) compared mortality rates, however, were based on SHS centers and by sex, with daily ciga- with the Oklahoma (10.7 years) and Da- categorization of the exposure that in- rette use among Arizona women as low as kota (9.8 years) centers. However, adjust- cluded gangrene or amputation. This is six per day. Thus, the effects of smoking ment for center had no effect on the less specific than our exposure, which in- on mortality may be less in American In- strength of association between LEA and cluded only documented amputation at- dians because of the lower dose of to- mortality. tributable to diabetes. Thus, it is not bacco. Although there is no evidence Foot ulceration and LEA in diabetic surprising that mortality risk estimates supporting the absence of negative health individuals occur most often as a result of from this report are higher than those re- effects of tobacco use in this cohort, these the combined effects of vascular and ported by the WHO Multinational Study. effects may be overwhelmed by a combi- nerve damage resulting from exposure to It would be interesting to examine the ef- nation of the strength of other risk factors a prolonged hyperglycemic environment, fect on mortality of vascular reconstruc- and the relatively low dose of tobacco pathways that are consistent with data tion before or after a lower-extremity among ever-smokers. presented in this report (29,30). Monofil- amputation. Unfortunately, these data are LEA is a risk factor for mortality prob- ament testing measures large nerve fiber not available in the SHS, making the effect ably because it is a highly specific marker function, an established risk factor for of these procedures on our risk estimates of damage to large and small vessels, as LEA (31,32). Although no data on neu- impossible to quantify with certainty. well as peripheral nerve damage resulting ropathy were collected at the baseline Although multivariable models atten- from long-standing diabetes. This hy- SHS exam, monofilament testing was uated the crude LEA effect on mortality, pothesis is supported by our data, which conducted at the second SHS examina- LEA remained a significant predictor of show longer duration of diabetes among tion, which occurred ϳ5 years after base- all-cause mortality in both men and individuals with an LEA, a high preva- line. These data showed that duration of women and of CVD mortality among lence of peripheral arterial disease and re- diabetes was significantly associated with women. These results suggest that the ef- nal dysfunction among individuals with a insensitivity to the monofilament (28), a fect of LEA on mortality was not entirely lower-extremity amputation, a relation of finding that is consistent with data from attributable to diabetes-related comor- these factors with mortality in our cohort, the current analysis showing an associa-

DIABETES CARE, VOLUME 27, NUMBER 6, JUNE 2004 1291 Amputation and mortality in American Indians tion between LEA and duration of diabe- The authors acknowledge the assistance 10. Lee JS, Lu M, Lee VS, Russell D, Bahr C, tes. and cooperation of the AkChin Tohono Lee ET: Lower-extremity amputation: in- We are unaware of previous studies O’odham (Papago)/Pima, Apache, Caddo, cidence, risk factors and mortality in the that have quantified the strength of asso- Cheyenne River Sioux, Comanche, Delaware, Oklahoma Indian Diabetes Study. Diabe- ciation between the anatomical site of Spirit Lake, Fort Sill Apache, Gila River Pima/ tes 42:876–882, 1993 Maricopa, Kiowa, Oglala Sioux, Salt River 11. Nelson RG, Gohdes DM, Everhart JE, LEA and subsequent mortality risk. The Pima/Maricopa, and Wichita Indian commu- Hartner JA, Zwemer FL, Pettitt DJ, adjusted HRs for all-cause mortality for nities, without whose support this study Knowler WC: Lower-extremity amputa- toe and BKA compared with those with- would not have been possible. The authors tions in NIDDM: 12-yr follow-up study in out a lower-extremity amputation were also thank the IHS hospitals and clinics at each Pima Indians. Diabetes Care 11:8–16, 2.0 and 2.3, respectively. Although sur- center and Betty Jarvis, RN, Tauqeer Ali, MD, 1988 vival analysis showed significant differ- PhD, and Marcia O’Leary, RN, directors of the 12. Rith-Najarian SJ, Valway SE, Gohdes DM: ences in survival between diabetic SHS clinics and their staffs. Dr. Chris Burd Diabetes in a northern Minnesota subjects with and without an LEA, the provided support for the study of foot disease Chippewa Tribe: prevalence and inci- graded survivorship among lower- in the SHS. dence of diabetes and incidence of major complications, 1986–1988. Diabetes Care extremity amputation sites did not reach 16:266–270, 1993 statistical significance. The lack of signif- References 13. Valway SE, Linkins RW, Gohdes DM: Ep- icant findings was due to limited statisti- 1. Lee ET, Keen H, Bennett PH, Fuller JH, Lu idemiology of LEAs in the Indian Health cal power associated with the small M: Follow-up of the WHO Multinational Service, 1982–1987. Diabetes Care number of site-specific LEAs. Study of Vascular Disease in Diabetes: 16:349–353, 1993 Participants who reported diabetes at general description and morbidity. Diabe- 14. Mayfield JA, Reiber GE, Nelson RG, baseline had an average diabetes duration tologia 44 (Suppl 2):S3–S13, 2001 Greene T: A foot risk classification system of 12 years. Diabetes duration was longer 2. Knowler WC, Saad MF, Pettitt DJ, Nelson to predict diabetic amputation in Pima In- in those with an LEA, and it increased as RG, Bennett PH: Determinants of diabetes dians. Diabetes Care 19:704–709, 1996 the site of amputation became more prox- mellitus in the Pima Indians. Diabetes 15. Lee ET, Welty TK, Fabsitz R, Cowan LD, Care 16:216–227, 1993 Lee NA, Oopik AJ, Cucchiara AJ, Savage imal. Participants with a toe amputation 3. Gohdes D: Diabetes in North American PJ, Howard BV: The Strong Heart Study: a had an average diabetes duration of 19 Indians and Alaska Natives. In Diabetes in study of cardiovascular disease in Ameri- years; participants with a BKA had an av- America. 2nd ed. Bethesda, MD, National can Indians: design and methods. Am J erage diabetes duration of 21 years, and Institute of Diabetes and Digestive and Epidemiol 132:1141–1155, 1990 the participant with an AKA had diabetes Kidney , National Institutes of 16. Howard BV, Welty TK, Fabsitz RR, for 34 years. These findings highlight the Health, 1995 (NIH publ. no. 95-1468) Cowan LD, Oopik AJ, Le NA, Yeh J, Sav- effect of duration of diabetes on severity of 4. Knowler WC, Bennett PH, Hamman RF, age PJ, Lee ET: Risk factors for coronary diabetic complications and on mortality Miller M: Diabetes incidence and preva- heart disease in diabetic and nondiabetic risk, and they are consistent with the lence in Pima Indians: a 19-fold greater Native Americans: the Strong Heart higher prevalence of LEA in Arizona, the incidence than in Rochester, Minnesota. Study. Diabetes 41 (Suppl. 2):4–11, 1992 Am J Epidemiol 108:497–505, 1978 17. Howard BV, Lee ET, Cowan LD, Fabsitz center with the longest mean duration of 5. Lee ET, Lu M, Bennett Ph, Keen H: Vas- RR, Howard WJ, Oopik AJ, Robbins DC, diabetes. cular disease in younger-onset diabetes: Savage PJ, Yeh JL, Welty TK: Coronary Increases in the occurrence of type 2 comparison of European, Asian, and heart disease prevalence and its relation to diabetes in American Indian children may American Indian cohorts of the WHO risk factors in American Indians: the “push back the clock” with respect to de- Multinational Study of Vascular Disease. Strong Heart Study. Am J Epidemiol 142: velopment of complications. Diabetic Diabetologia 44 (Suppl. 2):S78–S81, 2001 254–268, 1995 complications such as LEA are likely to 6. Chaturvedi N, Stevens LK, Fuller JH, Lee 18. Resnick HE, Valsania P, Phillips C: Diabe- begin appearing earlier in life because of ET, Lu M: Risk factors, ethnic differences tes and non-traumatic lower extremity the earlier onset of diabetes among today’s and mortality associated with LEA in dia- amputation in Black and White Ameri- youth. Without intensive prevention ef- betes: the WHO Multinational Study of cans: the NHANES I Epidemiologic Fol- Vascular Disease in Diabetes. Diabetologia low-up Study, 1971–1992. Arch Int Med forts, today’s children will spend a greater 44 (Suppl. 2):S65–S71, 2001 159:2470–2475, 1999 proportion of their adult lives coping with 7. Freeman WL, Hosey GM: Diabetic com- 19. Fabsitz RR, Sidawy AN, Go O, Lee ET, the social and physical difficulties associ- plications among American Indians of Welty TK, Devereux RB, Howard BV: ated with an LEA and other diabetes com- Washington, Oregon, and Idaho: preva- Prevalence of peripheral arterial disease plications. Our findings of excessive lence of retinopathy, end-stage renal dis- and associated risk factors in American mortality associated with amputation ease, and amputations. Diabetes Care 16: Indians: the Strong Heart Study. Am J Epi- suggest that future research should focus 357–360, 1993 demiol 149:330–338, 1999 on efficacy of interventions such as sur- 8. Wirth RB, Marfin AA, Grau DW, Helger- 20. Morgan C, Lazarow A: Immunoassay of veillance of foot health and patient self- son SD: Prevalence and risk factors for di- insulin: two antibody system: plasma in- care and education. abetes and diabetes-related amputations sulin levels in normal, subdiabetic and di- in American Indians in southern Arizona. abetic rats. Diabetes 12:115–126, 1963 Diabetes Care 16:354–356, 1993 21. Clauss A: [Rapid physiological coagula- 9. Farrell MA, Quiggins PA, Eller JD, Owle tion method in determination of fibrino- Acknowledgments— This study was sup- PA, Miner KM, Walkingstick ES: Preva- gen]. Acta Haematol 17:237–246, 1957 ported by grants U01-HL-41642, U01-HL- lence of diabetes and its complications in 22. Carter SA: The role of pressure measure- 41652, and U01-HL-41654 from the National the Eastern Band of Cherokee Indians. Di- ments in vascular disease. In Non-Invasive Heart, Lung, and Blood Institute. abetes Care 16:253–256, 1993 Diagnostic Techniques in Vascular Disease.

1292 DIABETES CARE, VOLUME 27, NUMBER 6, JUNE 2004 Resnick and Associates

Bernstein EF, Ed. St. Louis, MO, Mosby, 1998 oraro RF: The independent contributions 1985, p. 513–44 26. Acton K, Rogers B, Campbell G, Johnson of diabetic neuropathy and vasculopathy 23. Yao IST: Pressure measurement in the ex- C, Gohdes D: Prevalence of diagnosed di- in foot ulceration. How great are the risks? tremity. In Vascular Diagnosis. 4th ed. abetes and selected related conditions of Diabetes Care 18:216–219, 1995 Bernstein EF, Ed. St. Louis, MO, 1993, p. six reservations in Montana and Wyo- 30. Chaturvedi N, Abbott CA, Whalley A, 169–175 ming. Diabetes Care 16:263–265, 1993 Widdows P, Leggetter SY, Boulton AJ: 24. Resnick HE, Lindsay RS, McDermott MM, 27. Welty TK, Lee ET, Yeh J, Cowan LD, Go Risk of diabetes-related amputation in Devereux RB, Jones KL, Fabsitz RR, O, Fabsitz RR, Le NA, Oopik AJ, Robbins South Asians vs. Europeans in the UK. Howard BV: Relationship of high and low DC, Howard BV: Cardiovascular disease Diabet Med 19:99–104, 2002 ankle brachial index to all-cause and car- risk factors among American Indians: the 31. Adler AI, Boyko EJ, Ahroni JH, Smith DG: diovascular disease mortality. Circulation Strong Heart Study. Am J Epidemiol 142: Lower-extremity amputation in diabetes: 109:733–739, 2004 269–287, 1995 the independent effects of peripheral ar- 25. Lee ET, Cowan LD, Welty TK, Sievers M, 28. Sosenko JM, Sparling YH, Hu D, Welty T, terial disease, sensory neuropathy, and Howard WJ, Oopik A, Wang W, Yeh J, Howard BV, Lee E, Robbins DC: Use of foot ulcers. Diabetes Care 22:1029–1035, Devereux RB, Rhoades ER, Fabsitz RR, Go the Semmes-Weinstein monofilament in 1999 O, Howard BV: All-cause mortality and the Strong Heart Study: risk factors for 32. Boyko EJ, Ahroni JH, Stensel V, Forsberg cardiovascular disease mortality in three clinical neuropathy. Diabetes Care 22:1 RC, Davignon DR, Smith DG: A prospec- American Indian populations aged 45–74 715–1721, 1999 tive study of risk factors for diabetic foot years, 1984–1988: the Strong Heart 29. McNeely MJ, Boyko EJ, Ahroni JH, ulcer: the Seattle Diabetic Foot Study. Di- Study. Am J Epidemiol 147:995–1008, Stensel VL, Reiber GE, Smith DG, Pec- abetes Care 22:1036–1042, 1999

DIABETES CARE, VOLUME 27, NUMBER 6, JUNE 2004 1293