Visual Abstract
Abstract
Background and objectives Dyslipidemia, a risk factor for cardiovascular disease, is common in CKD but its change over time and how that change is influenced by concurrent progression of CKD have not been previously described.
Design, setting, participants, & measurements In the CKD in Children study we prospectively followed children with progressive CKD and utilized multivariable, linear mixed-effects models to quantify the longitudinal relationship between within-subject changes in lipid measures (HDL cholesterol, non-HDL cholesterol, triglycerides) and within-subject changes in GFR, proteinuria, and body mass index (BMI).
Results A total of 508 children (76% nonglomerular CKD, 24% glomerular CKD) had 2–6 lipid measurements each, with a median follow-up time of 4 (interquartile range [IQR], 2.1–6.0) years. Among children with nonglomerular CKD, dyslipidemia was common at baseline (35%) and increased significantly as children aged; 43% of children with glomerular CKD had dyslipidemia at baseline and demonstrated persistent levels as they aged. Longitudinal increases in proteinuria were independently associated with significant concomitant increases in non-HDL cholesterol (nonglomerular: 4.9 [IQR, 3.4–6.4] mg/dl; glomerular: 8.5 [IQR, 6.0–11.1] mg/dl) and triglycerides (nonglomerular: 3% [IQR, 0.8%–6%]; glomerular: 5% [IQR, 0.6%–9%]). Decreases in GFR over follow-up were significantly associated with concomitant decreases of HDL cholesterol in children with nonglomerular CKD (−1.2 mg/dl; IQR, −2.1 to −0.4 mg/dl) and increases of non-HDL cholesterol in children with glomerular CKD (3.9 mg/dl; IQR, 1.4–6.5 mg/dl). The effects of increased BMI also affected multiple lipid changes over time. Collectively, glomerular CKD displayed stronger, deleterious associations between within-subject change in non-HDL cholesterol (9 mg/dl versus 1.2 mg/dl; P<0.001) and triglycerides (14% versus 3%; P=0.004), and within-subject change in BMI; similar but quantitatively smaller differences between the two types of CKD were noted for associations of within-subject change in lipids to within-subject change in GFR and proteinuria.
Conclusions Dyslipidemia is a common and persistent complication in children with CKD and it worsens in proportion to declining GFR, worsening proteinuria, and increasing BMI.
- dyslipidemia
- pediatric nephrology
- lipids
- cardiovascular disease
- chronic kidney disease
- child
- humans
- HDL cholesterol
- triglycerides
- glomerular filtration rate
- risk factors
- body mass
- follow-up studies
- cholesterol
- proteinuria
- dyslipidemias
- HDL lipoproteins
- cardiovascular diseases
- chronic renal insufficiency
Introduction
Dyslipidemia is a well recognized atherosclerotic cardiovascular disease risk factor found commonly in children and adults with CKD. Our baseline cross-sectional study of the CKD in Children (CKiD) study cohort demonstrated a 45% prevalence of dyslipidemia, consistent with more familiar data from adult patients with CKD (1,2). The increased burden of adverse clinical events related to atherosclerotic risk is clear in adults with CKD, whereas the risk to children with CKD is most evident in a younger onset of atherosclerosis-related outcomes among young-adult survivors of childhood-onset CKD and ESKD (3–5). The latest dyslipidemia management guidelines for patients with CKD suggest clinical benefit in initiating lipid-lowering therapy in selected older or higher-risk adults with nondialysis CKD; there is no recommendation to treat younger adults with lower risk, or children (3–5). In particular, the recommendation not to treat dyslipidemia in children with CKD stems from a lack of sufficient data showing safety and benefit, and is consistent with recommendations for younger adults with a lower risk of cardiovascular events within 10 years. Moreover, these guidelines recommend long-term longitudinal research about lipids in children and young-adult survivors of childhood CKD to clarify the risks, outcomes, and unique characteristics of this group. Such information is important for developing a comprehensive understanding of cardiovascular risk in children with CKD, as well as designing studies of potential interventions to improve dyslipidemia in CKD.
Because cross-sectional data demonstrate a positive association between the degree of dyslipidemia and CKD severity manifesting as lower GFR and/or greater proteinuria (1,2,6), it seems logical that the progression of CKD would associate with worsening dyslipidemia longitudinally. However, there are no studies that have prospectively followed individuals, child or adult, with CKD over many years to verify if progression of CKD is associated with progression of dyslipidemia, the magnitude of the link, and which specific elements of CKD associate with specific lipid measures. We contrast this question to the frequently addressed theme of whether baseline dyslipidemia is linked to kidney outcomes such as incident or progressive CKD (7).
To address the gaps in longitudinal knowledge about lipids in children with CKD, our objective was to describe trajectories of lipid measures in children enrolled in the prospective, observational CKiD study. We hypothesized that individual declines in GFR and increases in proteinuria over time would associate with worsening dyslipidemia longitudinally.
Materials and Methods
CKiD is a multicenter, prospective, longitudinal cohort study in the United States and Canada, enrolling children aged 1–16 years with mild to moderate CKD. Its methodology has been previously described and the study protocol has been reviewed by each center’s institutional review board, with participants and parents providing written informed consent/assent according to local requirements (8). Briefly, children were recruited between 2004 and 2019 from 55 clinical sites; in-person follow-up visits occur annually.
The primary outcomes of interest were within-subject changes in fasting lipid levels—specifically, total cholesterol, HDL cholesterol, non-HDL cholesterol, and triglycerides—over follow-up. As part of the study protocol, total cholesterol, HDL cholesterol, and triglycerides are measured every other year, beginning with the first follow-up visit. Non-HDL cholesterol was calculated as the difference between total cholesterol and HDL cholesterol. This marker is considered of similar epidemiologic value to LDL cholesterol, does not require additional laboratory testing, captures other atherogenic lipoprotein classes such as “remnants,” which can be disproportionately elevated in CKD, and is more robust to nonfasting status as well (9–12). Additionally, dyslipidemia was defined at each patient-visit as having one or more of the following abnormal indicators: hypertriglyceridemia (triglycerides >130 mg/dl), decreased HDL cholesterol (<40 mg/dl), and elevated non-HDL cholesterol (>160 mg/dl). We chose these cut-off points on the following basis: (1) normative The National Health and Nutrition Examination Survey (NHANES) data available for children aged ≥12 years, (2) normative data of the Lipid Research Clinic data set, and (3) levels frequently considered atherogenic in adults (13–15). These cut-off levels are compatible with, or more conservative than (less likely to be classified as dyslipidemic), those suggested by more recent expert guidelines on pediatric dyslipidemia (16,17). Analysis was restricted to data from participants with at least two sets of lipid measurements. Baseline was defined as the first study visit at which lipids were measured (first follow-up visit). The timing of the repeated measures among the participants is shown in Supplemental Table 1.
The primary risk factors of interest were change over follow-up from baseline in GFR, proteinuria, and standardized body mass index (BMIz), respectively. These variables were defined at each visit as xij − xi0, where xij was the GFR, log-transformed proteinuria, or BMIz at the jth visit for the ith person, and xi0 was the corresponding baseline measurement for the ith person. At each visit, GFR was estimated using the CKiD-derived and published “bedside” estimating equation, (18) GFR=41.3×height/sCr, with height measured in meters and sCr indicating serum creatinine, measured in milligrams per deciliter using enzymatic methodology. Proteinuria was quantified by the first-morning urine protein-to-creatinine ratio in mg/mg (UPCR), as previously described (19). Age- and sex-specific body mass index (BMI) percentiles and standardized scores (BMIz) were calculated using 2000 Centers for Disease Control and Prevention standard growth charts for children in the United States. Overweight was defined as a BMI between the 85th and 95th percentiles, and obesity as a BMI over the 95th percentile (20).
Multivariable, linear mixed-effects models were used to quantify the longitudinal relationship between the respective lipid measures and GFR, UPCR, and BMI z-score. The models included the primary risk factors of interest and were adjusted for the following baseline variables: age, sex, black race, glomerular versus nonglomerular CKD diagnosis, GFR, log-transformed UPCR, and BMI z-score. Interaction terms of glomerular diagnosis and the three longitudinal change variables of interest allowed for separate estimates of these parameters for children with glomerular and nonglomerular CKD. The models also included a fixed-effect parameter for time from baseline (i.e., slope) as well as random-effects for both the intercept and slope. As such, the models captured both the between-individual differences in lipids as well as the within-individual lipid changes over follow-up as a function of GFR, UPCR, and BMI z-score. All parameters listed in the table were included simultaneously in the model. Thus, the primary relationships of interest, within-person changes in lipid measures, and their association with concomitant changes in GFR, UPCR, and BMI are adjusted for both baseline effects and follow-up time. As such, these parameters are estimated comparing children of similar attained age over follow-up. Triglycerides were modeled as log-transformed values to meet the assumptions of normality (21). As such, estimated differences and changes in triglycerides are reported as percentages. Additional methodologic details are provided in Supplemental Appendix 1.
Results
At the time of analysis, 891 children with a median GFR of 53 (interquartile range, 39–68) ml/min per 1.73 m2 had enrolled in the CKiD study, of whom 508 had baseline (study year 2, when lipids were first measured) plus at least one set of follow-up lipid measures. Collectively, these children had 1514 person-visits and 2199 years of follow-up. The median follow-up time was 4.0 (interquartile range, 2.1–6.0) years, with more than half (58%) of the children contributing three or more sets of lipid measures. Baseline study visit characteristics are shown in Table 1 for children with nonglomerular and glomerular CKD. On average, children with glomerular CKD were older, more frequently of black race, had shorter duration of CKD, higher GFR, greater proteinuria, higher BMI percentiles, and shorter follow-up in both years and total number of longitudinal lipid measures. Total cholesterol, HDL cholesterol, non-HDL cholesterol, and triglyceride levels, as well as the prevalence of dyslipidemia, were similar between the two diagnosis groups. A total of 16 children (3%) had non-HDL cholesterol >190 mg/dl at baseline. Use of lipid-lowering medication was infrequent (overall, 2%; n=9), although slightly higher in the glomerular CKD group.
Baseline characteristics by CKD diagnosis, n=508
Among the nonglomerular group, the cross-sectional prevalence of uncontrolled dyslipidemia increased with increasing age, from 31% to 49% (Cochran–Armitage trend test, P=0.001). As shown in Figure 1, this trend was accompanied by an increase in the proportion of children with nonglomerular CKD presenting with multiple elevated lipid measures. Similar trends were not evident in the glomerular CKD group. For the majority of children with glomerular CKD (those at least 11 years old), dyslipidemia prevalence appeared to be independent of age, although a small sample of younger children with glomerular CKD (n=34) displayed lower dyslipidemia prevalence. Lipid-lowering therapy use also increased significantly over time (P=0.01, data not shown), but remained very uncommon at just 3% (n=51) of all person-visits from 27 children.
Cross-sectional prevalence of dyslipidemia by age. Defined as one or more abnormality of triglycerides >130 mg/dL or HDL-cholesterol <40 mg/dL or non-HDL cholesterol >160 mg/dL.
Table 2 provides the parameter estimates from multivariable, linear mixed-effects models for each of the lipid measures of interest. From the study population, 18 patients and 211 person-visits were excluded because of missing covariate data, leaving 490 patients and 1303 person-visits for this adjusted analysis. The parameter estimates labeled “baseline level: between-person differences” capture the associations between baseline lipid levels and risk factors and are analogous to the results of a cross-sectional analysis. Parameter estimates for the primary risk factors of interest appear under the subheading “within-person change from baseline.” For both the nonglomerular and glomerular groups, longitudinal increases in UPCR and BMIz were independently associated with significant concomitant increases in non-HDL cholesterol and triglycerides. A doubling of the UPCR level over follow-up in children with nonglomerular CKD was associated with an expected increase in non-HDL cholesterol of 4.9 mg/dl (95% confidence interval [95% CI], 3.4 to 6.4) in the same time period; among children with glomerular CKD, the expected increase was greater, at 8.5 mg/dl (95% CI, 6.0 to 11.1; P value for heterogeneity=0.01). Likewise, a 0.5-unit increase in age- and sex-specific BMIz over follow-up was associated with an expected concomitant increase in the triglyceride level of 3% (95% CI, 1% to 5%) among children with nonglomerular CKD and 14% (95% CI, 7% to 21%) among children with glomerular CKD (P value for heterogeneity=0.004). With the exception of longitudinal decreases in GFR and longitudinal increases in BMIz in children with nonglomerular CKD, longitudinal changes in HDL cholesterol were not associated with concomitant changes in GFR, UPCR, or BMIz. Compared with changes in UPCR and BMIz, GFR change displayed weaker associations with changes in lipid levels. Observed patterns of change in total cholesterol closely mirrored those of non-HDL cholesterol. In nine out of the 12 (three risk factors and four lipid measurements) estimated associations of interest, children with glomerular CKD displayed stronger, negative point estimates than their nonglomerular CKD counterparts and five of these differences were statistically significant.
Results from multivariable, random-effects models of total cholesterol, HDL cholesterol, non-HDL cholesterol, and triglycerides on select demographic and clinical factors of interest
As a sensitivity analysis, we ran the models presented in Table 2, excluding those observations (n=51) at which lipid-lowering therapy was reported. The results were qualitatively and quantitatively similar. Additionally, adjusting for duration of CKD in the children with glomerular CKD for whom length of diagnosis is less than attained age did not affect the results of these models.
Finally, the parameter estimates under the subheading “slope” show rates of change (per 2 years) in lipid levels expected among children with no change in GFR, UPCR, and BMIz over follow-up. The results demonstrate, by comparison, that the most significant detrimental changes in lipids were associated with worsening in GFR, UPCR, and BMIz over follow-up (“within-person change from baseline”) and are not merely owing to passage of time since disease onset. Specifically, children presenting with stable (nonchanging) GFR, UPCR, and BMIz over follow-up had statistically significant expected (as seen in the general population at this median age [22]) decreases in total cholesterol (−2.1 mg/dl per 2 years; 95% CI, −3.9 to −0.2) and non-HDL cholesterol (−3.0 mg/dl per 2 years; 95% CI, −4.7 to −1.4) and significant increases in HDL cholesterol (−3.0 mg/dl per 2 years; 95% CI, 0.3 to 1.6) per 2 years. Triglycerides increased on average 5% (95% CI, 2% to 7%) per 2 years.
Discussion
This study demonstrates the degree to which change in kidney function and proteinuria are associated with concomitant dyslipidemic changes in children with CKD. Overall, worsening CKD is accompanied by worsening dyslipidemia. These associations are evident beyond the relationships attributable to baseline effects, are modified by disease type (glomerular versus nonglomerular CKD), and interact with change in BMI.To our knowledge, this study is the first specific examination of the relationship between within-subject longitudinal change in GFR and proteinuria to within-subject longitudinal change in lipids. Traditional methods for analyzing longitudinal data capture relationships between baseline exposures and longitudinal changes in outcome. Such approaches do not simultaneously account for changes in the exposure markers over follow-up, which is both common and expected in clinical practice. We believe that our analysis, capturing the association between longitudinal changes of both important risk factors (GFR, proteinuria, and BMIz) and lipid marker outcomes of interest, is novel and unique. Moreover, because dyslipidemia is an intermediate marker of cardiovascular risk, the ability to track these changes precisely over an extended period is an important step toward the ultimate goal of understanding how cardiovascular outcomes relate to earlier and variable degrees of exposure. Our results strengthen the concept that slowing CKD progression and preventing obesity should slow the development of dyslipidemia and, subsequently, cardiovascular disease. Moreover, in studying the effect of CKD on development of dyslipidemia in children, there was a comparatively low level of confounding. Unlike adults with CKD, most children have primary causes of CKD and do not commonly present with competing causes of dyslipidemia other than obesity (if present). In that respect, our cohort allows perhaps the best possible model of this pathophysiologic complication of CKD in humans, making our findings generalizable to other populations with CKD. The median age of our cohort was around 12 years, and in healthy children at that age, total cholesterol normally begins to fall over time (22) (which our data supports both at baseline and longitudinally). By way of contrast, this allowed additional opportunity to demonstrate the effects of CKD having an opposing effect.
The strength of the within-subject analysis rests in comparing each patient to themselves over time. Although we were able to demonstrate that the point prevalence of dyslipidemia increased over time in the cohort as a group, this finding is potentially confounded because of the changing construct of the cohort over time and subsequent changes in the cohort characteristics, such as the proportion of participants with glomerular disease, the average BMI percentile, and the median GFR. One reason for this is that patients requiring dialysis or transplant exit the study during follow-up. The within-subject analysis also provides a level of precision exceeding both the baseline analysis and cohort-level analysis, and minimizes confounding by allowing an individual to serve as their own control. Thus, it was possible to demonstrate that decreasing GFR affected HDL and non-HDL cholesterol, whereas worsening proteinuria affected non-HDL cholesterol and triglycerides; meanwhile, worsening BMI affected HDL and non-HDL cholesterol, and triglycerides. The success of this approach suggests that the use of within-subject exposure change analysis may yield additional information in other longitudinal cohorts as well. For example, it is interesting to note that in the Study of Heart and Renal Protection (SHARP) study, LDL cholesterol in the placebo group was overall nearly unchanged over 4 years, despite the fact that CKD was progressing (23). The relationship between within-subject changes in CKD and lipids was not reported in either the placebo or the treatment group. Our results suggest post hoc analyses considering individuals’ changes in GFR, proteinuria, and BMI over time might yield additional information from that study. Although our analysis additionally presents expected changes in the lipid measures of interest over time for children with stable (nonchanging) GFR, UPCR, and BMIz over follow-up, these results are limited in scope epidemiologically because the majority of children in the study have progressing CKD; nonetheless, these results largely match observations in the healthy childhood population at the same median baseline age (22).By demonstrating that CKD progression (defined as a decline in GFR or worsening proteinuria) is associated with worsening dyslipidemia, our study suggests that it remains important to investigate the underlying mechanisms as they are not fully understood. Recent findings support insulin resistance and increased ApoC-III as important factors in CKD-related dyslipidemia (24–26). Basic and translational research findings support additional factors such as reduced levels of lipoprotein lipase in target tissues, as well as abnormal levels of lipoprotein receptors (27). Such lines of investigation could be useful if they suggest CKD-specific mechanisms of disease that may have targeted avenues of treatment.
Although the focus of this study was the relationship between changes in GFR and proteinuria and change in lipids, it is also important to note that longitudinal increases in BMI were independently associated with concomitant dyslipidemic changes. It is notable that although these effects were observed for both glomerular and nonglomerular subgroups, with the exception of HDL cholesterol, the effect of BMI on lipids was much greater in the glomerular disease category (versus nonglomerular). We suspect that certain treatments for glomerular disease, especially steroids, may lead to both the obesity and dyslipidemia that are involved in both processes; this observed interaction merits further study. Regardless, obesity and its antecedents, such as low physical activity and high-calorie diets, must be highlighted as modifiable risk factors for dyslipidemic progression (28,29).
The current observation that worsening GFR and worsening dyslipidemia are associated is generally consistent with our previous cross-sectional finding that dyslipidemia was associated with lower GFR among the patients with nonglomerular CKD in our cohort (30). Nonetheless, dyslipidemia as a risk factor for CKD progression has been inconsistently demonstrated in other studies (31–34). In the Chronic Renal Insufficiency Cohort (CRIC) study of adults with CKD, there was no overall association. In fact, subgroup analysis in patients without significant proteinuria demonstrated that higher LDL cholesterol was associated with a 26% lower risk of the composite kidney end point (7). However, in the CRIC study, about one fifth of the patients had preexisting vascular disease, more than half were already receiving lipid-lowering agents, and more than half had diabetes. This again highlights a strength of our study, which is the opportunity to study the relationship between CKD progression and dyslipidemia in individuals with almost no additional confounding clinical factors. Nonetheless, our study is observational and we cannot infer causality.
Guidelines for lipid management of children with CKD specifically suggest no pharmacologic treatment (such as statins) and have suggested the need for further research to fill specific knowledge gaps (3). As a prospective, observational cohort study, our data does not clarify any threshold for treatment, but does address the gaps in knowledge; here, we demonstrate a persistently high prevalence of dyslipidemia as a cardiovascular risk exposure in children, which increases concomitantly with the progression of CKD. This strengthens the idea that disease duration should be considered in risk assessment and guideline development, and not just age. Further follow-up of the CKiD cohort may prove useful in that regard if future data allow us to link disease duration to eventual cardiovascular outcomes.
Despite the strengths of our study, there are also several weaknesses and analytic challenges. Loss of patients from the cohort because of progression to ESKD may present selection bias in that our analysis does not include as many patient-years of data for that subset of patients that progressed most quickly. However, if this bias exists, we will most likely be underestimating the strength of the association. Change in lipids during childhood, and particularly during puberty, is a known source of variation (35,36). We adjusted for age and sex in the analysis to account for this phenomenon, but residual confounding may remain. Although the relationships we explored and quantified may differ across age groups, we were not able to further stratify our analysis because of data size limitations. Also, although a subset of patients had a baseline measure of insulin resistance, it was not measured thereafter, so we were unable to analyze that relationship longitudinally. Finally, our study does not address surrogate markers of lipid-mediated vascular disease, such as intimal medial thickness or pulse-wave velocity.
In conclusion, dyslipidemia remains common among children with CKD over many years. Individuals with CKD develop, maintain, and experience worsening of dyslipidemia over time in proportion to the degree of baseline CKD and the progression of CKD over time. Declining GFR and worsening proteinuria are independently associated with worsening dyslipidemia. Within-subject change in GFR and proteinuria over time, in addition to baseline status, should be incorporated into future observational and interventional studies in patients with CKD.
Disclosures
Dr. Flynn, Dr. Furth, Dr. Kupferman, Dr. Mitsnefes, Dr. Pierce, Dr. Saland, and Dr. Warady have nothing to disclose.
Funding
The CKD in Children (CKiD) study is supported by grants from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK; U01-DK-66143 (to Dr. Warady), U01-DK-66174 (to Dr. Furth), U01-DK-082194, and U01-DK-66116), with additional funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development and the National Heart, Lung, and Blood Institute. Dr. Flynn is supported by a grant from NIDDK (U01-DK066143). Dr. Furth and Dr. Warady are supported by grants from the National Institutes of Healthhttps://doi.org/10.13039/100000002. Dr. Mitsnefes, Dr. Pierce, and Dr. Saland are supported by grants from the NIDDK related to the CKiD study.
Supplemental Material
This article contains the following supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.03110319/-/DCSupplemental.
Acknowledgments
We wish to thank the CKD in Children (CKiD) study participants and their families, as well as the site investigators, for many years of dedication to this project.
Data in this manuscript were collected by the CKiD prospective cohort study with clinical coordinating centers (principal investigators) at Children’s Mercy Hospital and the University of Missouri−Kansas City (Dr. Warady), Children’s Hospital of Philadelphia (Dr. Furth), Central Biochemistry Laboratory (Dr. George Schwartz) at the University of Rochester Medical Center, and data coordinating center (Dr. Alvaro Muñoz) at the Johns Hopkins Bloomberg School of Public Health. The CKiD website can be found at www.statepi.jhsph.edu/ckid.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
- Received March 13, 2019.
- Accepted October 9, 2019.
- Copyright © 2019 by the American Society of Nephrology
References
In this issue
Jump to section
More in this TOC Section
Original Articles
Chronic Kidney Disease
Cited By...
- No citing articles found.