Triglycerides and Amputation Risk in Patients With Diabetes: Ten-year follow-up in the DISTANCE study

Kaiser Permanente Medical Care Program (“Kaiser Permanente”) is a fully integrated, nonprofit, group practice health plan that provides comprehensive medical services to over ∼2.5 million members (January 1995) throughout Northern California (including the San Francisco Bay and Sacramento metropolitan areas) or ∼25–30% of the surrounding population. The Kaiser Permanente membership closely approximates the general population ethnically and socioeconomically except for the extreme tails of the income distribution (11).

In 1993, Kaiser Permanente established the Kaiser Permanente Northern California Diabetes Registry. The registry is updated annually by identifying all health plan members with diabetes from automated databases for pharmacy, laboratory, hospitalization records, and outpatient diagnoses (12–14). The registry sensitivity is estimated to be 96% based on chart review (15). During 1994–1996, all noninstitutionalized registry members 19 years or older identified as having diabetes prior to 1 January 1996 (n = 90,302) were selected for participation in a health survey (self-administered questionnaire) or a computer-assisted telephone interview in English or Spanish. Eighty-three percent of eligible members participated in the survey. After removing those denying having diabetes or discontinuing membership, 70,748 respondents with diabetes remained. The survey collected information on, among other things, ethnicity, information needed to classify diabetes type (see Karter et al. for algorithm [16]), family history of diabetes, education, and behavioral risk factors. Survey respondents were ethnically diverse: 14% black, 12% Asian, 10% Latino, and 64% non-Latino White. Of those self-identifying as “Asian,” 44% were Filipino, 24% were Chinese, 12% were Japanese, and 19% were Korean, other Asian, or mixed race. We compared the demographic composition (age, sex, socioeconomic status) of diabetes survey responders to nonresponders in a previous study (14) and found no evidence to suggest responder bias. In addition to survey-derived data, we obtained measures of neighborhood-level socioeconomic status by linking each member’s address to the 1990 census block group-level average annual 1989 per capita income.

We used a survey follow-up study design to evaluate the incidence density of LEA and its predictors, restricting analysis to the 28,701 survey respondents who had triglyceride data within 1 year of baseline. Initial triglyceride levels were measured after overnight fasting. Baseline was defined as 1 January 1995 for all members surveyed in 1994 and the survey date (January 1995 to March 1997) for those surveyed after that date. The duration of follow-up (person-time) was tabulated through membership records and ended with an event censoring due to dropping of Kaiser Permanente membership for any period greater than 2 months, death, or the end of the study (31 December 2006). We excluded those with a history of prior LEA events noted in hospital discharge records during the 5 years prior to baseline. Prognostic, confounding, and stratifying variables (see below) were ascertained at or prior to baseline from automated records, the diabetes survey, and the 1990 census.

Nontraumatic LEA procedures were identified from discharge codes (ICD-9-CM procedure codes 84.10–84.17). We included patients with any amputation of the lower extremity. In two separate validations of amputation procedures (A. Karter, personal communication, n = 109; and J.V. Selby and D. Zhang [17], n = 209), 99% of electronic hospital discharge records were confirmed by chart review.

Using direct standardization with the entire diabetes cohort as the population standard, we calculated age-adjusted sex- and ethnicity-specific rates for LEA. Proportional hazards regression (Cox) models were then used to calculate adjusted hazard ratios (HRs) as an estimate of the relative risk of elevated triglycerides (<150, 150–199, 200–499, and >500 mg/dL) associated with LEA. After detecting no violations of the proportionality assumption, we specified a series of Cox regression models, including a base demographic model (age-, race-, and sex-adjusted plus an indicator for elevated triglycerides, LDL, and HDL) and specified saturated models that added socioeconomic variables (race, individual-level education, average census block level income, and proportion in working class occupations), health behaviors (smoking status, alcohol intake, self-reported treatments for diabetes including diet and exercise, and frequency of self-monitoring of blood glucose), and clinical factors including type of diabetes, diabetes duration, type of diabetic medications, first-degree family history of diabetes, BMI (underweight, normal, overweight, obese) (18), height quartiles (19,20), hypertension (self-reported and pharmacy records of antihypertensive medication dispensing), peripheral neuropathy (self-report) (17,21), A1C (<7, 7–8, 8–10, >10 g/dL), LDL (<100, 100–129, 130–159, >160 mg/dL), HDL (<40, 40–59, >60 mg/dL), and use of statin or other lipid-lowering agents.

Patients who sustained an LEA were older (61 years [SD 11.2] vs. 59 years [11.2]), had a longer duration of diabetes (14 years [9.7] vs. 9 years [9.5]), and had a higher hemoglobin A1C (9.05 g/dL [2.12] vs. 8.36 g/dL [1.94]) compared with those who did not sustain an LEA (Table 1). The age- and sex-adjusted nontraumatic LEA incidence rate (95% CI) was 2.3 (1.5–3.1) per 1,000 person-years (Table 2). The sex-specific, age-adjusted rates (95% CI) were 1.4 (1.2–1.7) per 1,000 person-years for women and 3.2 (1.6–4.8) for men. The age- and sex-adjusted incidence rates (95% CI) for white, black, Hispanic, and Asians were 2.7 (1.5–4.0), 2.0 (1.6–2.5), 1.4 (1.1–1.8), and 0.9 (0.5–1.3), respectively.

Crude HRs for triglyceride, LDL, and HDL levels revealed a significant, stepwise relationship with triglycerides: 150–199 mg/dL, HR 1.10 (95% CI 0.92–1.32); 200–499 mg/dL, 1.27 (1.10–1.47); and >500 mg/dL, 1.65 (1.30–2.10) (reference <150 mg/dL). There was no significant association for LDL or HDL with amputation risk. Minimally adjusted Cox proportional hazards models including triglyceride, LDL, and HDL levels continued to demonstrate a stepwise association between triglycerides and LEA risk: 150–199 mg/dL, HR 1.07 (95% CI 0.89–1.29); 200–499 mg/dL, 1.20 (1.03–1.40); and >500 mg/dL, 1.39 (1.03–1.88) (reference <150 mg/dL). LDL was again not significantly associated with amputation risk, whereas HDL showed a protective effect between 40–59 mg/dL (HR 0.74 [95% CI 0.64–0.87]) but not at >60 mg/dL (1.21 [0.92–1.60]) (reference HDL <40 mg/dL). A further-adjusted Cox proportional hazards model (model 1 in Table 3) including triglycerides, LDL, HDL, age, sex, and race also revealed a stepwise association of triglycerides and LEA risk.

The stepwise association between triglycerides and LEA persisted despite further adjustment for socioeconomic (individual-level education, average census block level income, and proportion in working class occupations), health behavioral (smoking, BMI, drinking, adherence to guidelines for self-monitoring of blood glucose, and exercise), and clinical variables (statin medication, fibrates/niacin medication, family history of diabetes, height, duration of diabetes, A1C, type of diabetes and therapy, hypertension, neuropathy, retinopathy, stroke, heart attack, and end-stage renal disease [ESRD]) (model 2 in Table 3). As expected, duration of diabetes, A1C, height, hypertension, neuropathy, retinopathy, ESRD, stroke, and heart attack were also associated with amputation. Interestingly, statin medication use and fibrates/niacin medication use did not have statistically significant HRs. LDL >160 mg/dL and HDL >60 mg/dL were the only levels of these tests associated with an increased risk of amputation; it appeared to be a threshold effect for LDL, but there was no consistent or stepwise relationship for HDL.

Additionally, the rates of elevated triglycerides did differ substantially by race (21, 40, 46, and 46% among African Americans, Asians, Whites, and Hispanics, respectively; P < 0.001). However, we failed to detect a significant interaction between triglyceride level and race (P = 0.83), suggesting that the relationship between triglycerides and amputation risk did not differ by race. Furthermore, similar results were seen with triglyceride level and amputation risk using the entire diabetes cohort and utilizing a missing indicator for those who did not have triglyceride levels at baseline. We also found that, as expected, there was a significant, inverse correlation between triglyceride and HDL levels (correlation = −0.24, P < 0.0001). However, this modest collinearity was not considered sufficient to distort our triglyceride effect estimates from the Cox regression models.

In this large, multiethnic cohort of diabetic patients, we found that elevated triglyceride levels are associated with LEA even after adjusting for a host of potential confounders. The consistency of the relationship between triglycerides and LEA across ethnic groups was further supportive. All models revealed a stepwise increase in amputation risk with increasing triglyceride levels. This is in contrast to low HDL, which was not associated with increased amputation risk. The threshold of LDL levels above 160 mg/dL was also associated with increased risk of amputation. Dyslipidemia has been proposed as a potential risk factor in the development of diabetes complications such as retinopathy and neuropathy based on results utilizing the DCCT/EDIC cohort and the EURODIAB study (7,22). Our results suggest that high triglyceride levels, independent of the other major lipid components, may put patients at risk for one major diabetes complication—LEA. Whether this association applies more widely to other vascular diabetes complications is unknown.

These results add to the finding in prior studies that elevated triglyceride levels may be a risk factor for complications of diabetes (5,7,8), including LEA (9,10). However, the current study involved a larger population and found the association to be statistically significant even after adjusting the models for a multitude of variables that were not assessed in the prior studies.

Given the current state of the literature, the guidelines on triglyceride management do not advocate aggressive treatment. The National Cholesterol Education Program recommends diet and exercise for patients with triglyceride levels between 150–199 mg/dL, to consider treatment in high-risk patients with levels 200–499 mg/dL, and to use pharmacologic treatment when levels are 500 mg/dL or greater to prevent pancreatitis, but not necessarily to prevent microvascular and macrovascular complications (23). Clearly, further studies are needed to ascertain the role of hypertriglyceridemia in these diabetic sequelae. Our study gives further support to the notion that triglycerides may be one of the key modifiable risk factors in the development of amputations. If this association holds in clinical trials, then clinicians may have a target with the potential to improve important patient outcomes.

One of the main limitations of this study is that a large proportion of our population did not have data on triglyceride, HDL, and LDL levels during the year prior to baseline; at the time, these assessments were not done nearly as frequently or routinely as is common today. However, our sensitivity analysis using missing triglycerides as an indicator and evaluating the entire diabetes cohort suggests that missing data did not substantially bias our findings. Also, observational cohort studies, particularly cross-sectional designs, can make erroneous assumptions between cause and effect when suggestive variables correlate. In our study, the longitudinal study design and exclusion of patients with prevalent amputations at baseline precludes time-ordering violations (i.e., LEA preceding the triglyceride ascertainment). Another limitation is that observational studies are vulnerable to residual confounding; however, we were able to adjust for a wide array of confounders, which should greatly reduce this concern.

The major strengths of this study are its longitudinal cohort design, validated and complete capture of LEA events, and large sample size with rich array of clinical and behavioral data. Additionally, our diabetes registry captures patients who are not treated with diabetes medications (medical nutritional therapy) through laboratory findings and outpatient diagnosis records, providing a more representative diabetic sample than pharmacy-based registries that only capture patients receiving prescriptions or studies from diabetes specialty clinics that typically include patients with more severe disease.

In summary, elevated triglyceride level was associated with subsequent LEA independently of the other lipid components and a wide range of potential confounders in this large, well-characterized diabetic cohort. Though specific guidelines exist for cholesterol (LDL and HDL levels) management in this patient population, only vague guidelines exist for triglyceride management (23). This observational study suggests that triglyceride levels are predictive of amputation risk in a stepwise fashion. Despite standards of glycemic, blood pressure, and cholesterol control, patients continue to develop microvascular and macrovascular complications (combination of neuropathy and macroangiopathy is thought to cause LEAs), and triglyceride control may be an important additional primary prevention effort. More research is necessary to define a causal role of triglyceride levels on amputation risk in diabetic patients. Based on this current robust cohort study, such research should be a priority.

This study was supported by the National Institutes of Health (NIDDK R01-DK065664 and DK081796) and by the Neurology Training Grant Fellowship (T32).

The Program for Neurology Research & Discovery was instrumental in providing salary support for B.C.C.

No potential conflicts of interest relevant to this article were reported.

B.C.C. was involved in interpretation of the statistical analysis and wrote the manuscript. E.F. proposed the hypothesis and helped with the manuscript. J.L. performed the statistical analysis. K.K. helped with statistical interpretation and contributed to the manuscript. R.P.-B. contributed to the manuscript. H.M. reviewed the manuscript. A.J.K. was integrally involved in the creation of the cohort and was significantly involved in analysis and writing the manuscript.

Parts of this study were presented in abstract form at the American Neurologic Association meeting, San Francisco, California, 12–15 September 2010.