|Home | About | Journals | Submit | Contact Us | Français|
Cystatin C has been proposed to better estimate renal function and predict cardiovascular disease (CVD) than serum creatinine. To expand on our previous report, we investigated whether the relationship of cystatin C to progression of coronary artery atherosclerosis (CA) differed between individuals with type 1 diabetes (T1D) and persons without diabetes.
Coronary artery calcium was measured twice over 2.4±0.4 years (n=1,123, age=39±9 years, 47% male, 45% T1D). Significant CA progression was defined as a ≥2.5 increase in square root calcium volume score or development of clinical coronary artery disease. Stepwise multiple logistic regression was performed to investigate whether the association of cystatin C to CA progression differed by T1D status.
The main finding and novelty of this article is that while the univariate association of cystatin C to CA progression was similar in T1D patients and persons without diabetes mellitus and in the expected direction (increased cystatin C as a biomarker of worsening renal function associated with CA progression), the association of cystatin C to progression of CA differed by T1D status (P=0.01) after adjustment for other CVD risk factors. Unexpectedly, in persons without diabetes mellitus having relatively normal renal function, increased cystatin C was associated with decreased CA progression (odd ratio [OR]=0.65, 95% confidence interval 0.44–0.96, P=0.029) after adjustment, primarily due to adjustment for body mass index (BMI). Removal of BMI from this model resulted in a 49% change in the OR.
Our hypothesis-generating data suggest a complex relationship among cystatin C, BMI, and CA progression that requires further study.
Extensive literature exists on renal disease as a cardiovascular disease (CVD) risk factor in persons with and without type 1 diabetes (T1D).1–5 Data suggest that cystatin C may be a better estimator of renal function than serum creatinine.6–11 Cystatin C was less influenced by age, sex, and race than serum creatinine as an estimate of glomerular filtration rate (GFR) in a large cohort with chronic kidney disease12 and in longitudinal data in young adults with T1D and normal renal function at baseline.11 Cystatin C has been proposed to better estimate the slope in decline in GFR than serum creatinine and therefore to better detect trends in change in GFR to allow for clinical intervention.13,14
Numerous epidemiological investigations have reported that cystatin C as a biomarker of renal function better predicts CVD events and death than serum creatinine (or serum creatinine-based estimates of GFR).15–19 Consistent with these studies, we reported that in T1D subjects in the Coronary Artery Calcification in Type 1 Diabetes (CACTI) Study, elevated cystatin C (indicating worsening GFR) was associated with an increased risk of progression of coronary artery atherosclerosis (CA).20
Therefore, our aim was to expand on our previous report in only individuals with T1D20 to investigate whether differences exist between T1D patients and subjects without diabetes mellitus (DM) (who had relatively normal renal function) in the relationship of cystatin C to progression of CA in the CACTI Study cohort.
The data presented in this report were collected as part of the baseline examination of 1,416 participants in the CACTI Study who were 19–56 years of age and included 652 participants with T1D and 764 participants without DM.21 All participants were asymptomatic for coronary artery disease (CAD) and had no history of coronary artery bypass graft, coronary angioplasty, or unstable angina. Participants with diabetes generally had been diagnosed when <30 years of age or had positive antibodies or a clinical course consistent with T1D. Of the 1,416 persons enrolled at baseline, 1,355 (93%) had available stored serum to measure cystatin C. Of these, 1,123 (83%) had data on progression of CA (mean follow-up time, 2.5±0.4 years). Participants who did not have follow-up data were younger at baseline than those with follow-up data. All participants provided informed consent, and the study was approved by the Colorado Combined Institutional Review Board.
Participants completed the baseline examination between March 2000 and April 2002, and a more detailed description of the study and baseline characteristics of this cohort including collection of anthropometric, demographic, and laboratory data has been published.22 Insulin resistance was approximated as the inverse of the estimated glucose disposal rate.23 GFR (or creatinine clearance) was estimated by both the Cockcroft-Gault formula (GFRCG)24 and the Modification of Diet in Renal Disease equation (GFRMDRD),25 both based on serum creatinine. We also used the Mayo Clinic (GFRMC) equation.26
After an overnight fast, blood was collected, centrifuged, and separated. Serum was stored at −70°C until assayed. Cystatin C was measured on stored serum samples in the clinical laboratory at University of Colorado Hospital, Denver, CO, using a commercially available particle enhanced immunonephelometric assay (Dade-Behring, Newark, DE). Stored samples from the subjects' baseline study visit had previously been thawed once. The coefficient of variation was 3.3%. Intra-assay precision is 2.3–4.1%, and inter-assay precision is 2.6–3.3%, per the package insert. Results are reported in mg/L with a sensitivity cutoff of 0.23mg/L.
All participants underwent two electron beam tomography scans using an Imatron (South San Francisco, CA) C-150 Ultrafast CT scanner within 5min without contrast at baseline and two scans at follow-up as previously described22 using the standard acquisition protocol.27 The volume scores were calculated using the volumetric method, which is based on isotropic interpolation.28
In this study, we defined CAC progression as reported by Hokanson,29 who noted that bias in the interscan variability of calcium volume score (CVS) values exists such that the variability increases as levels of coronary calcium increase; a difference between baseline and follow-up square root transformed CVS ≥2.5 is used to signify significant change in CVS because a change of this magnitude is <1% likely to be due to interscan variability. In addition, participants who had a CAD event (myocardial infarction, coronary artery bypass graft, angioplasty with stent, or death attributed to CAD as adjudicated by a three physician committee) were also considered as having CA progression based on having had a CAD event.
Data are presented as arithmetic means and SDs for continuous variables (geometric means and ranges for log-transformed variables) and percentages for categorical variables stratified by T1D and CA progression status (see Table 1). To investigate whether the association of cystatin C to CA progression differed by T1D status, stepwise multiple logistic regression in a combined model was performed including a cystatin C×T1D interaction term with a value of P<0.1 as the criteria for entry and removal from the model. Age, baseline CVS, and sex were forced into all models. All variables in Table 1 were considered for inclusion in the stepwise model predicting CA progression. This model is presented with variables added sequentially: first, demographics (age, sex) and baseline CVS; second, body mass index (BMI), systolic blood pressure (SBP), and high-density lipoprotein (HDL), all CVD risk factors that entered the stepwise model; and, finally, to account for possible differences by T1D status, T1D interactions with all variables if their P values were <0.1 (see Table 2). Next, stepwise multiple logistic regression was performed in a model in the subjects with DM only, and each variable was removed to assess its role as a confounder (see Table 3). Cystatin C was also examined by quartiles, and its association to CA progression was approximately linear. Odds ratio (OR) values are presented per SD change in the predictor variables.
Because covariate adjustment had different effects in the non-DM model versus the T1D model, we next examined the correlation between cystatin C and each of the covariates stratified by diabetes status. Fisher Z transformations were used to test for significant differences between correlation coefficients for T1D patients versus subjects without DM (see Table 4). Covariates with a correlation coefficient value of P<0.1 for either T1D patients or subjects without DM were entered into backward elimination linear regression models stratified by T1D status. Variables with P<0.05 in either stratified model were then forced into both models, and Wald tests were used to compare estimates from the two models (i.e., to test for interaction). We used stratified models for these analyses to allow the residual variance to differ by T1D status. Beta coefficients for each covariate standardized to the entire cohort and their respective P values were calculated for the stratified models (see Table 5).
Baseline characteristics are presented in Table 1 with differences between the T1D and non-DM groups as previously reported.30 Within the T1D and non-DM groups, differences existed by progression status with those who progressed having elevated CVD risk factors. Cystatin C was higher and had greater variance in T1D patients than in persons without DM (0.84±0.37mg/L vs. 0.78±0.10mg/L, P=0.0006; P<0.0001 for equality of variance). Of note is that cystatin C was only minimally (but statistically significantly) different by progression status in subjects without DM (0.81±0.10mg/L vs. 0.78±0.10mg/L; P=0.02), but albumin excretion rate (AER) did not differ. In contrast, renal function markers were significantly different by progression status in the T1D subjects. Also, BMI was similar in T1D progressors and non-progressors but much higher in non-DM progressors than in non-progressors adjusted for sex (data not shown).
The relationship between cystatin C and CA progression was explored in stepwise multiple logistic regression in the full cohort (Table 2 and Fig. 1). Results of this analysis demonstrated the expected association of increasing cystatin C to CA progression in T1D participants, even with sequential adjustment for additional CVD risk factors as previously reported in an analysis including only T1D subjects.20 In subjects without DM, cystatin C had the expected association to CA progression in univariate analysis (Model 1, OR=1.33, 95% confidence interval [CI] 1.04–1.69, P=0.029), but after adjustment for age, sex, and baseline CVS this association approached the null and was not significant (Model 2, OR=1.03, 95% CI 0.78–1.38, P=0.82). With further adjustment for covariates that entered the stepwise model (Model 3) and then with significant interaction terms added (Model 4), the association between cystatin C and CA progression differed significantly by T1D status (interaction P values: P=0.03 for Model 3, P=0.01 for Model 4). In the final adjusted model (Model 4), cystatin C was positively associated with CA progression in individuals with T1D (OR=1.13, 95% CI 1.03–1.26, P=0.01) and inversely associated in persons without DM (OR=0.73, 95% CI 0.52–1.02, P=0.067), although only reaching borderline significance in the non-DM model.
Given these unexpected results and to further investigate the relationship of cystatin C to CA progression in participants without DM, the relationship of cystatin C to CA progression was explored in additional stepwise multiple logistic regression analysis including the subset of subjects without DM with AER data (n=553). In participants without DM, cystatin C was significantly associated with less CA progression (OR=0.65, 95% CI 0.44–0.96, P=0.029) while adjusting for other covariates that entered the model (BMI, which entered the model first, then cystatin C, then AER) in addition to covariates forced into the model (age, sex, and baseline CVS) (Table 3). Each SD increase of cystatin C (0.10mg/L) was associated with a 35% decrease in the odds of CA progression. Similarly, for each SD decrease in AER the odds of CA progression increased by 32%. AER appeared to be missing at random in our dataset (missingness was not related to any of the variables of interest in our study), and repeating the model with a missing value indicator showed the same results. Next, confounders of the relationship of cystatin C to CA progression were investigated, and removal of BMI from the model presented in Table 3 resulted in the largest change toward the null in the OR for cystatin C (49%) and AER (32%), respectively, indicating BMI is the primary confounder of the association of cystatin C (and AER) to CA progression. Similar results were obtained with waist circumference and visceral fat in place of BMI. (Removal of other covariates resulted in changes between 2% and 17% [data not shown].) Additionally, a cystatin C×BMI interaction term was entered into the model in Table 3, but it was not significant (P=0.89).
To further investigate why covariate adjustment may have had different effects in the non-DM model versus the T1D model, correlation coefficients between cystatin C and other variables, stratified by T1D status, were determined (Table 4). In T1D patients and subjects without DM, renal-related measures were correlated with cystatin C, whereas obesity-related measures were significantly correlated in persons without DM. Homocysteine was correlated with cystatin C in both T1D patients and those without DM. Variables with significantly different correlation (P<0.05) in T1D as compared to persons without DM included gender, BMI, waist circumference, visceral fat, subcutaneous fat, SBP, HDL, plasminogen activator inhibitor-1 (PAI-1), fibrinogen, serum creatinine, AER, GFRCG, GFRMDRD, and homocysteine.
Correlates of cystatin C stratified by T1D status were evaluated with multiple linear regression analysis with a backward elimination approach (Table 5). Stratified models were then fit to the data including variables significant in either group, and Wald tests were used to compare estimates from the two models (i.e., test for interaction). In both T1D patients and persons without DM, gender, HDL, BMI, fibrinogen, serum creatinine, and AER were significantly associated with cystatin C. Additionally, PAI-1 and homocysteine were significantly associated with cystatin C in persons without DM. Significant interactions by T1D status existed for gender, BMI, fibrinogen, serum creatinine, and AER with stronger correlations in patients with T1D.
The main finding and novelty of this article is that although the univariate association of cystatin C to CA progression was similar in T1D patients and persons without DM and in the expected direction (increased cystatin C as a biomarker of worsening renal function associated with CA progression), the association of cystatin C to progression of CA differed by T1D status after adjustment for other CVD risk factors in multivariate analysis (P=0.01 for the T1D×cystatin C interaction). Unexpectedly, in persons without DM increased cystatin C was associated with decreased CA progression after adjustment, primarily driven by adjustment for BMI. In T1D patients, the measured risk factors did not explain all of the association between cystatin C and progression of CA, and the association remained significant after adjustment.
This is the first article, to our knowledge, investigating the association of cystatin C to CA progression in a relatively healthy cohort without DM with normal renal function and the first to examine wither the association differs by T1D status. Because the findings from this study were unexpected and counterintuitive, it should be considered a hypothesis-generating study requiring confirmation in additional datasets. Cystatin C has been proposed to be a superior marker of GFR than serum creatinine or serum creatinine-based estimates of GFR14,31 and to be a better predictor of CVD deaths15 and progression of CA.20 Given these reports in the literature, cystatin C may be used more commonly both in large epidemiological studies and clinically. Therefore, our observation on the differing relationship in multivariate analysis between T1D patients and persons without DM of cystatin C to CA progression is important for future applications and merits further explanation and investigation.
A number of possible explanations exist to explain these data. Once adjusted for BMI, elevated cystatin C has a protective effect on CA progression (P=0.029) in those persons without DM with relatively normal renal function and less variability in cystatin C than the T1D subjects. Of note is that the association of AER and serum creatinine (common markers of renal function) to cystatin C differed by T1D status, with serum creatinine less positively and AER inversely associated with cystatin C in individuals without DM compared to T1D patients. An intriguing possibility is that cystatin C could have a protective effect on the progress of atherosclerosis as has been noted in mouse and human models.32–38 In the CACTI cohort only cystatin C, and not serum creatinine, GFRCG, or GFRMDRD, was associated with CA progression. Although cystatin C has been used as a biomarker of GFR in epidemiological studies, basic science investigations have demonstrated that cystatin C inhibits elastin-degrading cysteine proteases, thereby diminishing their degradation of extracellular matrix proteins which is central in the atherosclerotic process.33 Furthermore, cystatin C gene polymorphisms were related to cystatin C concentrations39,40 and associated with more coronary artery stenoses39 but not with secondary CVD events.40 The possibility exists that although cystatin C serves as a reliable biomarker of GFR, it also certainly has a biologic function, and in persons with normal renal function higher concentrations of cystatin C could be protective in the atherosclerotic process, whereas in persons with impaired renal function these proposed beneficial biologic functions of cystatin C could be overwhelmed by the deleterious effects of kidney disease on CVD. Another possibility could be the complexity of the relationship of cystatin C to renal function and renal function to CAC. The Multi-Ethnic Study of Atherosclerosis has reported that cystatin C was not associated with cross-sectional CAC severity after adjustment for other risk factors in a population with mild to moderate kidney dysfunction.41
BMI has been reported to be a stronger predictor of CAC progression in individuals without DM than T1D persons in the CACTI study.42 In persons without DM cystatin C, as a marker of early renal dysfunction, could be on the causal pathway between increased BMI and CAD.43 In T1D persons higher BMI has been associated with CAC, but its relationship with CAC severity was inverse or nonexistent.44 In T1D patients, the relationship of glycemia to BMI (and other CVD risk factors) may not be as strong as the association of glycemia to CAD. In these data cystatin C was associated with renal function in both T1D patients and subjects without DM (although more strongly in T1D patients than in those without DM), whereas BMI (and other obesity measures) was associated with cystatin C in those without DM but not T1D patients.
Alternately, residual confounding may have been induced with adjustment for BMI. Other variables not measured or known to be measured imprecisely (e.g., physical activity and diet variables) may be causally associated with both BMI and CA progression. Failure to adjust for these additional risk factors could result in a model that overestimates the effect on BMI on CA progression, thus underestimating the cystatin C to CA progression association. There may also be other confounders (not necessarily associated with BMI) that were not measured or adequately controlled for in the current analysis.
Some limitations in our data need to be acknowledged and addressed in future studies. First, we do not compare cystatin C values to gold standard measures of GFR. However, recent data suggest cystatin C better estimates GFR than serum creatinine in subjects with chronic kidney disease,12 as do longitudinal data in young adults with T1D.11 Also, we use a surrogate marker of CAD instead of health outcomes such as CAD events or death as the CACTI cohort is relatively young and asymptomatic for CAD at enrollment, and the few CAD events as of this writing is a limitation (n=15 for the cohort [11 with T1D and four without DM] in this analysis; a sensitivity analysis performed after removing these subjects with CAD events did not change the results); data on patient outcomes are being collected prospectively. However, extensive methodologic detail has been taken to carefully define CAC progression in the CACTI cohort,29 and this methodology has been used in other datasets.45,46 Additionally, CAC has been used as a surrogate marker of CAD, with higher CAC indicating higher CVD risk in populations with a wide range of renal function.47
In conclusion, our data suggest a complex relationship between cystatin C and CA progression in subjects without DM having normal renal function with BMI as a significant confounder of the observed univariate association. In these data cystatin C was associated with renal function markers in both T1D patients and subjects without DM (although more strongly in T1D patients than in subjects without DM), whereas BMI (and other obesity measures) was associated with cystatin C in those without DM, but not T1D patients. As cystatin C is being more commonly used as a biomarker of renal function and a predictor of CVD, future studies should be aware of this relationship and interpret results accordingly. Longitudinal and genetic data on cystatin C as well as the relationship of cystatin C to CVD outcomes in additional cohorts will confirm or reject these relationships.
Support for this study was provided by NHLBI grant R01 HL61753, HL79611, and DERC Clinical Investigation Core P30 DK57516 from the National Institutes of Health. The study was performed at the Adult GCRC at the University of Colorado Denver Health Sciences Center supported by grant M01-RR00051 from the National Institutes of Health, at the Barbara Davis Center for Childhood Diabetes, and at Colorado Heart Imaging Center in Denver, CO. D.M.M. was supported by grant K23 DK075360. The results presented in this article have not been published previously in whole or part, except in abstract format.
No competing financial interests exist.