Serum triglycerides and mortality risk across stages of chronic kidney disease in 2 million U.S. veterans.

BACKGROUND: In the general population, elevated triglyceride (TG) levels are an important risk factor for cardiovascular disease and mortality. However, in chronic kidney disease, the association of serum TGs with mortality is less clear. OBJECTIVE: We sought to examine the association of TGs with mortality across chronic kidney disease (CKD) stages in a large cohort of U


Introduction
2][3] Mendelian randomization studies further corroborate a potential causal relationship between circulating TGs and atherogenesis by demonstrating an association linking genes determining higher TG levels and coronary heart disease (CHD), 4 and thereby indicate that TGs should be considered as an important cardiovascular risk factor and potential treatment target.A prior study showed that TG levels are elevated in patients with chronic kidney disease (CKD) and increase with advancing disease. 57][8] However, despite the abundant evidence linking CKD to TG dysmetabolism, it is still unclear how elevated TG levels in CKD impact outcomes and translate to cardiovascular endpoints in this population.A better understanding of this relationship may lead to insight of how to address elevated TGs in patients with CKD.
Patients with CKD have an elevated risk of mortality, in which over half of deaths are attributed to cardiovascular disease. 9It is known that both all-cause and cardiovascular mortality risks increase with advancing CKD. 10 Yet, it is unknown if TG levels are a contributor to the higher all-cause and cardiovascular mortality risks observed across all advancing CKD stages.Prior studies examining the relationship between TGs and mortality outcomes in patients with CKD have shown conflicting results, [11][12][13] and one study suggested that the TG-mortality relationship in CKD may be age dependent. 115][16] However, associations of TGs with mortality have not been evaluated across incrementally advancing CKD stages in a single large cohort.Thus, we hypothesize that the association of elevated TGs with mortality is impacted by advancing stages of CKD.

Study population and data source
LIPROVET (Lipid profiles and management in veterans with CKD) is a retrospective cohort study derived from administrative data sourced by the United States Veterans Affairs (VA) databases.It is composed of all veteran patients who received at least one serum lipid (TG or high-density lipoprotein [HDL], low-density lipoprotein [LDL], or total cholesterol) measurement between the baseline period of October 1, 2004 and September 30, 2006.Patients were followed until December 31, 2014.In the present study, patients were further excluded for missing a TG measurement during baseline, missing an estimated glomerular filtration rate (eGFR) measured within 90 days before the TG measurement, and missing information on censoring.Our final cohort comprised 2,086,904 veteran patients with a TG measurement (Fig. S1).
This study has been approved by the institutional review board of the Tibor Rubin VA Medical Center in Long Beach, CA.The required written consent was waived because of the large sample size, patient anonymity, and nonintrusive nature of the research.

Demographics and clinical measurements
All baseline clinical characteristics were extracted from the combination of VA and Centers for Medicare and Medicaid Services (CMS) databases, with additional supplementation from the United States Renal Data System (USRDS) databases.VA databases solely provided data on marital and smoking status. 17ipid-modulating therapies were predominantly extracted from VA pharmacy records with supplementation from CMS Medicare Part D, using specific drug class codes and names for classification.Statin or nonstatin therapy was defined as having the specific prescription at the time of the TG measurement.
Comorbidity at the time of the TG measurement, including the Deyo Charlson Comorbidity Index (CCI), was derived from combined VA and CMS data sets, using a 2 outpatient or 1 inpatient algorithm of International Classification of Diseases, Ninth Revision (ICD-9) Diagnostic and Current Procedural Terminology codes. 18,19ICD-9 codes were guided by those included in the Deyo CCI calculation, CMS chronic conditions, and prior studies. 18,20,21aboratory measurements, including the lipid panel, were obtained from the VA Managerial Cost Accounting System Laboratory Results.LDL was also calculated using the Friedewald 22 equation from other lipid measurements taken on the same day, only among patients initially missing an LDL measurement.Other laboratory measurements, including serum creatinine, were obtained from the VA Corporate Data Warehouse LabChem file.The Chronic Kidney Disease Epidemiology Collaboration formula was used to calculate eGFR, which was categorized into CKD stages (non-CKD, 3A, 3B, 4, and 5) at the time of TG measurement, according to Kidney Disease Improving Global Outcomes guidelines.23 Patients identified as with end-stage renal disease (ESRD) on renal replacement therapy according to the USRDS records at the time of TG measurement were classified as CKD stage 5, irrespective of eGFR level.Finally, data on body mass index (BMI) and blood pressure were obtained from the VA Corporate Data Warehouse Vital Signs file.24 For all analyses, the closest single measurement within 90 days before the TG measurement was used.

Exposure measurement
The primary exposure was a single measurement of TGs, categorized into the following groups: ,80, 80 to ,120, 120 to ,160 (reference), 160 to ,200, 200 to ,240, and $ 240 mg/dL, based on the distribution of the cohort, and clinically relevant cut points.

Outcome assessment
The primary outcomes were all-cause and cardiovascular mortality.Censoring for death and lost to follow-up were extracted from VA, National Death Index, CMS, and USRDS data sets.Lost to follow-up was determined by the last date of active use of VA or CMS services (inpatient, outpatient, or pharmacy).Cause of death was obtained from the National Death Index only and was categorized by specific ICD-10 codes for cardiovascular reasons of death (Table S1).Patients were followed up from the date of TG measurement to death, lost to follow-up, or December 31, 2014, whichever occurred first.

Statistical analysis
Patient characteristics were presented as mean (6standard deviation), median (interquartile range [IQR]), or proportion, where appropriate, and across TG groups and CKD stages.Multinomial logistic regression models were used to examine the odds of low TGs (,120 mg/dL) or high TGs ($240 mg/dL) vs moderate TGs (120 to ,240 mg/dL, reference).
Cox proportional hazard models were used to examine the association of TGs with all-cause or cardiovascular mortality stratified by CKD stage.For each CKD stage, TGs 120 to ,160 mg/dL was used as the reference.
For our analyses, 4 models were used with hierarchical adjustments: (1) unadjusted, (2) age adjusted, which included age, (3) case-mix adjusted, which included age, gender, race, ethnicity, and the following comorbid conditions: smoking status, CCI, myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, dementia, liver disease, cancer, diabetes, atrial fibrillation, hypertension, depression and ischemic heart disease, use of statin therapy, use of nonstatin lipidlowering drug therapy, and (4) case-mix 1 lab adjusted, which included the case-mix covariates as well as additionally adjusted for BMI and albumin and we identified as our fully adjusted model.In sensitivity analysis, we further adjusted for HDL and LDL in our case-mix 1 lab-adjusted model.Adjusted Wald's tests were performed to evaluate the interaction between TGs and CKD stage.
We also used restricted cubic splines to examine the association of continuous TGs with mortality within CKD stage, with best placed knots at the 5 th , 35 th , 65 th , and 95 th percentile of TGs per CKD stage.In subgroup analyses, we examined the TGs and mortality association across CKD stage within strata of age and diabetes comorbidity, as well as in males only.In addition, using restricted cubic splines with best-placed knots for eGFR, we evaluated the odds of high or low TGs compared with moderate TGs across continuous eGFR among those not with ESRD (n 5 2,074,884) and evaluated the mortality risk associations of low vs moderate TGs and high vs moderate TGs across continuous eGFR in a NDD-CKD (eGFR,60 mL/min/1.73m 2 , n 5 496,067) cohort.
Data were missing for ,1.4% and 4.2% of the baseline cohort for demographics and smoking status, respectively, and were imputed using a missing category.Baseline albumin and BMI data were missing for 27% and 11%, respectively, and were imputed using means.Data for other laboratory markers were missing at a similar rate.All analyses were performed using SAS Enterprise Guide (7.1) (Cary, NC) and Stata/MP Version 14 (College Station, TX).
In the total cohort, patients with higher TGs tended to be younger, white, and with a lower prevalence of chronic pulmonary obstructive disorder and anemia, yet a higher proportion of diabetes, depression, post-traumatic stress disorder, current smokers, and nonstatin users.Patients with higher TGs also had a greater BMI and total cholesterol, yet lower HDL.

Association of serum triglycerides with all-cause and cardiovascular mortality
A total of 726,992 all-cause and 246,530 cardiovascularrelated deaths occurred over a median [IQR] follow-up of 9.2 [6.5, 9.9] years for a crude rate of 44. 4 [44.3,44.5] allcause deaths and 15.1 [15.0, 15.1] cardiovascular deaths per 1000 person-years.Crude all-cause and cardiovascular death rates increased with advancing CKD stages (Table S4).In case-mix adjusted models, the association of TGs with all-cause and cardiovascular mortality appeared to be reverse J-shaped for non-CKD, stage 3A, 3B, and 4 patients (Tables S5 and S6).The relationship then assumed a more U-shaped association after additional adjustment for laboratory covariates.However, the strength of the association between high TGs ($240 mg/dL) with all-cause and cardiovascular mortality risk appeared to decrease across  worsening stages of CKD (P-interaction: ,0.0001, Fig. 1, Tables S5 and S6).Notably, among CKD stage 5/ESRD patients, TGs $ 240 mg/dL were associated with a lower risk of mortality compared to TGs 120 to ,160 mg/dL in the age and case-mix adjusted model, but after additional adjustment for BMI and albumin, the high TG-mortality relationship was null in these CKD stage 5/ESRD patients.Conversely, low TGs , 80 mg/dL were consistently associated with a higher risk of all-cause and cardiovascular mortality across CKD stages and all models of adjustment.Results were similar as illustrated by restricted cubic splines examining the association of continuous TGs with mortality outcomes in case-mix 1 lab-adjusted models, where the risk of mortality with higher TGs was null among CKD stage 5/ESRD patients (Figs. 2 and 3).Furthermore, among patients with NDD-CKD, the effect of worsening eGFR on the association of high TGs vs moderate TGs with all-cause and cardiovascular mortality is demonstrated in Figure S3B and S3D, where the strength of the associations decreased around an eGFR of 40 mL/min/1.73m 2 .The associations of low vs moderate TGs with all-cause and cardiovascular mortality risk were slightly higher across continuous levels of eGFR (Fig. S3A and S3C).

Subgroup analyses
Associations of TGs with all-cause and cardiovascular mortality across CKD stage were similar in strata of diabetes and age (,65 vs $ 65 years) for all-cause (Table S7) and cardiovascular mortality (Table S8).Although, for CKD stage 5/ESRD, low TGs (,80 mg/dL) were associated with a higher mortality risk in patients ,65 years; however, there was no difference in mortality risk for low TGs vs the referent in patients aged $65 years.Similar associations were evident when examining only male veterans.

Discussion
In a large national cohort of veteran patients, we observed that patients with CKD had a greater odds of having higher TGs ($240 mg/dL) independent of BMI, age, and comorbidities compared with moderate TGs (120 to ,240 mg/dL).We also observed that high TGs $ 240 mg/dL levels were associated with a higher risk of all-cause and cardiovascular mortality among non-CKD, stage 3A, 3B, and 4 patients, where the relationship was attenuated among CKD stage 5 and ESRD patients in models adjusted for demographics, comorbidities, lipid-altering therapies, BMI, and albumin.Low TGs , 80 mg/dL levels were also associated with a higher risk of all-cause and cardiovascular mortality across CKD stages.These relationships were consistent under restricted cubic splines analyses and across sensitivity analyses including stratification by age and diabetes, as well as among male veterans.Elevated TG levels in patients with CKD compared with non-CKD patients have been demonstrated in prior studies. 25,26In a previous cohort study, more advanced CKD stages were associated with a higher odds of hypertriglyceridemia defined as TGs $ 150 mg/dL 5 , where CKD stage 4 or 5 patients had a 2.5 times higher odds of hypertriglyceridemia compared with CKD stage 1.In addition, mechanisms responsible for dysregulation of lipoprotein metabolism leading to hypertriglyceridemia in CKD have previously been characterized. 7,8,27These mechanisms suggest that a major cause of hypertriglyceridemia in CKD is due to the deficient activity of enzymes involved in the metabolism of the TG-rich lipoproteins, thereby allowing a longer circulatory life span for these particles.In our study, odds of high TGs, defined as TGs $ 240 mg/dL, were also higher in advanced CKD stages compared with non-CKD patients in models adjusted for demographics, comorbidities, and laboratory markers of nutritional status.
Despite elevated TGs among patients with CKD, studies investigating the associations between TGs and mortality or cardiovascular outcomes in CKD have had conflicting results and were also limited by small size or residual confounding.Kovesdy et al. 12 found no association between TG quartiles and all-cause or cardiovascular mortality risk in a cohort of 986 NDD-CKD male veterans with models adjusted for case-mix variables plus surrogates of the malnutrition-inflammation-cachexia syndrome.Similar findings were observed in a community-based cohort of 1249 elderly patients with CKD. 28Chawla et al. also examined TG-mortality associations in 840 younger (mean age: 52 6 12 years) NDD-CKD patients with fewer comorbidities (5% diabetics) 13 and found no difference in all-cause and cardiovascular mortality risk across TG tertiles.
Alternatively, another small cohort of 807 patients with CKD from the Atherosclerosis Risk in Communities Study showed a positive association between TGs and a composite CHD outcome, although they did not find differences in association across CKD stage. 29The CKD stage stratified analysis, however, only adjusted for four demographic variables and therefore may be subject to residual confounding, as associations of log TG level with the CHD outcome in the study were attenuated after additional adjustment for comorbid conditions.In our study, higher TG levels were associated with both all-cause and cardiovascular mortality in patients with CKD even after adjustment for a number of potential confounders.In addition, we found that higher TG and mortality associations were modified and incrementally attenuated across worsening CKD stages.
In our study, we also found no effect modification by age group in associations of TGs with mortality across CKD 3A-4 stages.Our findings contrast a prior study by Navaneethan et al. 11 who examined TG-mortality associations in 25,641 Cleveland Clinic NDD-CKD (stages 3 and 4) patients.The authors found TGs were associated with all-cause mortality in patients younger than 65 years but not in older patients ($65 years old).However, their findings in their younger patients may have been driven by the lower mortality risk observed for patients with lower TGs.While in our cohort, lower TGs were associated with higher all-cause and cardiovascular mortality risk in both younger and older NDD-CKD stage 3 and 4 patients.A trend toward higher all-cause mortality risk for lower TGs has also been observed in a prior veteran cohort. 12It should be noted that younger VA patients may not be representative of the general population younger than 65 years, such as those treated at the Cleveland Clinic.These younger VA patients are eligible for VA health care due to military service, which may have led to the development of conditions not commonly present in the general younger male population.Therefore, lower TGs observed in our younger patients may represent malnutrition rather than a healthy lipid profile, thus leading to the higher observed mortality rates for lower TGs in our study.
In our cohort, higher TGs were associated with lower mortality risk in CKD stage 5/ESRD patients, which included a majority already on renal replacement therapy (n 5 12,020), but the relationship was attenuated to the null in models with adjustment for laboratory measurements including albumin, an important marker of malnutrition and inflammation. 30,31Low TG levels were associated with higher mortality risk in younger CKD stage 5/ESRD patients; however, there was no association between low TGs and mortality present for older CKD stage 5/ESRD patients.
Prior studies of hemodialysis patients have also shown that higher TGs trended toward 14,15 or were associated 16 with a paradoxically lower mortality risk in case-mixadjusted models, whereas lower TGs were associated with a higher mortality risk.However, associations for higher TGs were attenuated toward the null in models adjusted for malnutrition-inflammation-cachexia syndrome covariates. 14Liu et al. 31 hypothesized that malnutrition and inflammation in patients with ESRD may explain this inverse association between lipid markers and mortality in dialysis patients.While they also showed an inverse association between cholesterol and mortality in U.S. dialysis patients overall, they reported a positive cholesterol-mortality relationship in 189 dialysis patients without malnutrition and inflammation.Another study 32 similarly found that higher cholesterol levels were associated with a higher death risk in a subgroup of 128 Japanese dialysis patients with albumin $4.5 g/dL but not in subgroups with lower albumin levels.Conversely, some authors have criticized that these analyses based on small subsets of patients without inflammation or malnutrition may not be generalizable to all dialysis patients, who are typically afflicted with these conditions. 33,34Previous studies have not observed effect modification on TGmortality associations in dialysis patients on the basis on age; however, older patients with ESRD may have a higher prevalence of malnutrition and frailty overall, and therefore low TGs may no longer be as strong an indicator of malnutrition in consideration of other malnutrition factors such as albumin and BMI.
Although we adjusted for albumin, residual confounding by other markers of malnutrition and inflammation may still exist.Although we were unable to fully account for other markers because of high missingness, this possible residual confounding may explain the incrementally lower to null risk of mortality observed for higher TGs across advancing CKD stages.The underlying pathology explaining the reduced risk of mortality for higher TGs in patients with CKD is still unclear.In addition to potential complications due to malnutrition and inflammation, another possible explanation for this lack of relationship may be due to competing events of cardiovascular causes unrelated to atherosclerosis or TG-related cardiovascular disease, such as cardiomyopathies, left ventricular hypertrophies, or small vessel coronary disease.Tonnelli et al. 35 similarly found an attenuated relationship between LDL and myocardial infarction across worsening CKD stages in 836,060 adults and also postulated that this attenuation may be due to higher risk of cardiovascular outcomes due to malnutrition, inflammation, or competing cardiovascular causes unrelated to lipid levels.However, further studies are needed to assess these hypotheses.
7][38][39] Associations of high TGs with mortality outcomes were present although incrementally attenuated in patients with CKD stage 4 or earlier.In spline models showing continuous effect modification by eGFR in NDD-CKD patients on the association of high TGs with mortality, associations began to attenuate slightly around eGFR 40 mL/min/1.73m 2 .This may be explained by the restriction to a smaller subset of patients with moderate or high eGFR or by the placement of knots in the model.Nonetheless, this study has a number of strengths.It is one of the largest studies to investigate the association of TGs and mortality across CKD stages, especially among late CKD stages.Moreover, we were able to adjust for a number of potential confounders including smoking status and use of lipid-modulating therapies because of the wealth of our combined electronic medical records data sets.
However, a number of limitations should be noted for our analysis.Due to the observational nature of the study design, we cannot completely eliminate residual confounding nor make causal inferences on the relationship between TGs and mortality by CKD stage.Moreover, we adjusted for only available confounders, yet we were unable to fully account for other potential confounders such as other nutritional or inflammation markers such as C-reactive protein, dietary intake of saturated fat, abdominal adiposity, and alcohol intake.We assume that our TG measurements were drawn after fasting, given that most lipids were drawn in the morning, although we cannot confirm fasting status and a degree of misclassification remains possible.1][42] Finally, the VA population is primarily composed of older white males, and thus, our findings may not be generalizable to the general population, especially among females who may have differences in lipid metabolism. 43n conclusion, we observed that high serum TGs were associated with a higher risk of all-cause and cardiovascular mortality among non-CKD, stage 3A, 3B, and 4 patients, however, not among CKD stage 5/ESRD patients.Further studies are needed to examine the mechanism behind these relationships and to better understand how therapies aimed at lowering TGs may impact outcomes among patients with CKD.Data were presented as a poster at the 2017 American Heart Association Scientific Sessions meeting in Anaheim, California, November 14th, 2017 (ES).

Acknowledgments
Authors' contributions: E.S., K.K-Z., and C.P.K. contributed to research idea and study design and were responsible for supervision or mentorship; E.S. and M.S. contributed to data acquisition and statistical analysis; M.S., H.M., Y.O., C.P.K., K.K-Z., and E.S. contributed to data analysis/interpretation.Each author contributed important intellectual content during manuscript drafting or revision and accepts accountability for the overall work by ensuring that questions pertaining to the accuracy or integrity of any portion of the work are appropriately investigated and resolved.E.S. takes responsibility that this study has been reported honestly, accurately, and transparently that no important aspects of the study have been omitted and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

Figure 1
Figure 1 Association of serum triglycerides with (A) all-cause and (B) cardiovascular mortality stratified by CKD stage after casemix 1 lab adjustment.CKD, chronic kidney disease; ESRD, end-stage renal disease.

Figure 2
Figure 2 Restricted cubic splines of serum triglycerides with all-cause mortality across CKD stages after case-mix 1 lab adjustment.CKD, chronic kidney disease; ESRD, end-stage renal disease.

Figure 3
Figure 3 Restricted cubic splines of serum triglycerides with cardiovascular mortality across CKD stages after case-mix 1 lab adjustment.CKD, chronic kidney disease; ESRD, end-stage renal disease.

Table 1
Baseline characteristics of 2,086,904 patients stratified by serum triglycerides level This study was supported by Dr. Streja's Career Development Award IK2-CX 001266-01 from the US Veterans Administration Clinical Sciences Research and Development Program.H.M., C.P.K., K.K.Z., and E.S. are employees of the US Department of Veterans Affairs.The data reported here have been supplied by the US Veterans Administration.Support for VA/CMS data is provided by the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development, Health Services Research and Development, VA Information Resource Center (Project Numbers SDR 02-237 and 98-004).Opinions expressed in this presentation are those of the authors and do not represent the official opinion of the US Department of Veterans Affairs or the US Government.
Charlson Comorbidity Index, myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, dementia, liver disease, cancer, diabetes, atrial fibrillation, hypertension, depression, ischemic heart disease, use of statin therapy, and use of nonstatin lipid-lowering drug therapy.Case-mix 1 lab adjusted: case-mix adjusted and body mass index and albumin.Supplemental TableS5Association of serum triglycerides with all-cause mortality across levels of adjustment and CKD stage adjusted: age, gender, race, ethnicity, smoking status, Charlson Comorbidity Index, myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, dementia, liver disease, cancer, diabetes, atrial fibrillation, hypertension, depression, ischemic heart disease, use of statin therapy, and use of nonstatin lipid-lowering drug therapy.Supplemental TableS6Association of serum triglycerides with cardiovascular mortality across levels of adjustment and CKD stage Charlson Comorbidity Index, myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, dementia, liver disease, cancer, diabetes, atrial fibrillation, hypertension, depression, ischemic heart disease, use of statin therapy, and use of nonstatin lipid-lowering drug therapy.Case-mix 1 lab adjusted: case-mix adjusted and body mass index and albumin.Case-mix 1 lab adjusted 1 HDL 1 LDL: case-mix 1 lab adjusted 1 HDL 1 LDL.Supplemental TableS7Association of serum triglycerides with all-cause mortality across strata of age and diabetes and male gender under case-mix 1 lab adjustment Charlson Comorbidity Index, myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, dementia, liver disease, cancer, diabetes, atrial fibrillation, hypertension, depression, ischemic heart disease, use of statin therapy, use of nonstatin lipid-lowering drug therapy, body mass index, and albumin.Supplemental TableS8Association of serum triglycerides with cardiovascular mortality across strata of age and diabetes and male gender under case-mix 1 lab adjustment Model adjustments Case-mix 1 lab adjusted: age, gender, race, ethnicity, smoking status, Charlson Comorbidity Index, myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, dementia, liver disease, cancer, diabetes, atrial fibrillation, hypertension, depression, ischemic heart disease, use of statin therapy, use of nonstatin lipid-lowering drug therapy, body mass index, and albumin.