2008 USRDS Annual Data Report
This Chapter
Download PDF: Win*   Mac*
Download all slides: Win*   Mac*
*corresponding data in Excel included
Search This Page
Search All

One: Chronic kidney disease in the nhanes population

Assessing the burden of chronic kidney disease in the general population requires a random sample to ensure that the data are representative. In this chapter we use the dataset from the National Health and Nutrition Examination Survey (NHANES), produced by the National Center for Health Statistics at the CDC. To assess the burden of CKD across the country we apply published approaches based on various inclusion and exclusion criteria and on several equations for calculating the glomerular filtration rate — the basis of the CKD classification system. We first show the inter-relationship among diabetes, cardiovascular disease, hypertension, and CKD. Growth in the prevalence of CKD based on an eGFR of <60 ml/min/1.73 m2 (reported by Coresh et al. in November, 2007) is evident among participants age 60 and older, and among those with a diagnosis of diabetes. Data on the occurrence of microalbuminuria equal to or greater than 30 mg/g, in contrast, show less change. When the criteria are combined based on the CKD stages — using microalbuminuria in Stages 1–2 and an eGFR <60 for Stages 3–5 — growth appears to be centered within the population with a diagnosis of diabetes.

We also show the sensitivity and specificity, based on receiver operator curves, of several threshold values for abnormalities related to metabolic syndrome. These abnormalities generally give a low yield when simple criteria such as an eGFR <60 or microalbuminuria are used. They may need to be addressed at later stages of CKD, when the yield, along with the false positives and negatives, would be less.

Any assessment of CKD predictors is complex, since there is considerable interaction of CKD with race and age, and with other diseases such as diabetes, hypertension, and heart disease. We here examine traditional predictor analyses and also, in a more striking illustration, use a classification tree analysis, based on a recursive regression approach, to show disease interactions. This demonstrates the complexity of addressing CKD and of developing targeted detection programs to inform providers and the public about those who are at high risk. Based on this approach, it appears the major target groups should be patients age 65 and older, along with those younger than 65 with either diabetes or hypertension. Other risk factors add little to the percent of CKD patients detected, and may simply increase the cost of detection efforts rather than making them more successful.

The burden of comorbid conditions increases with more advanced stages of CKD. Cross-sectional data here illustrate this in part, but survival bias should also be considered. These data are consistent with reports on increasing rates of mortality and hospitalization presented by Go et al. (NEJM, 2004) and with the recent Lancet article from Wen et al. on a Taiwan population.

Awareness, treatment, and control of major risk factors for CKD are typical public health concerns when determining areas for potential improvement in this high-risk population. Data here suggest that, while awareness of the major disease such as diabetes, hypertension, and lipid disorders is high, control of factors such as blood pressure, lipid levels, and fasting blood sugar levels worsens with advancing CKD. Among those with CKD of Stages 4–5, for example, 57 percent have systolic blood pressures greater than or equal to 140 mmHg, and 74 percent have systolic pressures greater than 130 mmHg. Anemia is common in those with kidney disease, and, not surprisingly, increases with CKD stage. Data on iron status show modest degrees of lower transferrin saturation levels. And the NHANES dataset also includes information on the erythrocyte protoporphyrin level, another measure of iron deficiency, and a test of ineffective iron incorporation into the porphyrin ring. The percent of participants with elevated levels increases with more advanced CKD stages, and ranges from 11 percent in the non-CKD population to 56 percent in those with CKD of Stages 4–5.

Overall, data on the burden of CKD in the United States — as defined by the NHANES population sample — show that an increasing percentage of the population has evidence of CKD, and that there has been an increase of 20–25 percent from the 1988–1994 period. The CKD population carries a high burden of disease, and treatment and control lag behind recommended practices. Given the high rates of cardiovascular events, mortality, and progression to ESRD, target detection programs to improve care in high-risk populations may be reasonable. In Chapters Two, Three, and Four of this volume we present data on event rates, laboratory testing for risk factor monitoring, and treatment with medications known to reduce cardiovascular event rates and the progression of CKD, particularly in those with more advanced stages of the disease.

figure 1.1
In U.S. adults, chronic kidney disease (CKD) is more prevalent than either congestive heart failure or diabetes, and the prevalence of CKD appears to have grown over a period of approximately fifteen years. figure 1.2 An eGFR of <60 is especially prevalent among adults age 60 or older. figure 1.3 Overall, the prevalence of albuminuria has remained comparatively static over time, being present in 8.7 percent of the U.S. population in 1998–1994 and 9.3 percent in 2003–2006. table 1.g While 80 percent of adults in CKD stages 4–5 have hypertension, only 20 percent are being adequately managed in the sense of being aware of the abnormality and on a successful treatment regime.


figures 1.2, 1.3, 1.4, & 1.5 The prevalence of eGFR <60 ml/min/1.73 m2 has grown substantially, from 5.7 percent in 1988–1994 to 8.1 percent in 2003–2006, with most of the increase taking place between 1988 and 2002. This eGFR is especially prevalent among adults age 60 or older and those with diabetes or hypertension, and appears to have become more common among these older adults and in diabetics. The prevalence of albuminuria has remained comparatively static over time, present in 8.7 percent of the population in 1988–1994 and 9.3 percent in 2003–2006. Albuminuria is especially prevalent among adults age 60 or older, African Americans, those with hypertension, and, most notably, those with diabetes, though in the latter group its preva-lence has declined from 35.5 percent in 1988–1994 to 30.8 percent in 2003–2006. Prevalence of the composite disorder rose from 12.4 percent in 1988–1994 to 15.0 percent in 2003–2006, and from 31.9 to 37.8 percent in those age 60 or older.
Many treatment guidelines for patients with CKD suggest screening for metabolic compli-cations when eGFR levels are lower than 60 or when an abnormal albumin excretion is found. Figure 1.5 examines the diagnostic test performance of eGFR <60 and an abnormal urinary albumin-creatinine ratio for identifying typical metabolic abnormalities of CKD. In general, these strategies are highly insensitive for identifying the target metabolic disorders. Stated dif-ferently, this strategy would fail to identify most cases when metabolic competitions are present, and suggests that other screening strategies need to be developed.

table 1.a This table compares the prevalence of different stages of chronic kidney disease when alternative case definitions are applied to the overall population, as well as to major demographic and clinical subgroups. It is apparent that defining an abnormal urinary albumin-creatinine based on gender-specific criteria has a pronounced effect on the prevalence of Stage 1 and 2 CKD. It is also apparent that inclusion or exclusion of pregnant females has very little effect on the distribution of CKD stage among U.S. adults. A GFR of less than 60 ml/min/1.73 m2 is notably prevalent among adults age 60 or older (28.1 percent with Method 1) and subjects with diabetes (22.1 percent), hypertension (18.3 percent), and cardiovascular disease (31.7 percent).

table 1.b Here we show a multivariate model which reports adjusted odds ratios of CKD among U.S. adults. Dominant associations of CKD include age 60 or older (odds ratio 5.89), diabetes (odds ratio 2.50), cardiovascular disease (odds ratio 1.92) and hypertension (odds ratio 1.77).

table 1.c This table illustrates how different combinations of these latter variables perform as screening tests for identifying subjects with CKD. While all combinations are imperfect, the best combi-nation of sensitivity and specificity was obtained when the criteria for CKD-screening were age 60 or older or age <60 with diabetes or hypertension.

figure 1.6 Data show that a shift to the left in GFR distribution has occurred between NHANES III and the most recent NHANES populations. The greatest proportion of this change has occurred at GFR levels greater than 60 ml/min/1.73 m2.

figure 1.7 We show here a classification tree created using Classification and Regression Tree (CART) analysis, which relates predictor variables to a binary outcome — in this case, CKD. An advantage of this method over logistic regression is that the resulting tree shows interactions without having to pre-specify them. The most important predictors are higher on the tree; the most important predictor, therefore, is age, with a cutpoint of 64.5 giving the best separation between CKD and non-CKD. Within the younger age branch of the tree (age 20–64), the next most important split is diabetes status, followed by hypertension, a further age split, and obesity. Within the older age branch (age 65 and older), the next most important split is a further age split, followed by diabetes, and then another age split. Generally, the model shows that the most important factors are age, diabetes, hypertension, and obesity.

table 1.d & figure 1.8 Diabetes, hypertension, cardiovascular disease, and smoking are all more prevalent in participants with CKD Stages 1 –5 than in those with no CKD. For obesity, a non-stepwise pattern is present as the observed prevalence is greater in Stage 1 and lower in Stage 2 and 3. The association between stage of CKD and self-reported hypertension is especially striking and increases stepwise, such that 84.1 percent of participants in Stage 4–5 have hypertension.

figure 1.9 Cardiovascular disease is highly associated with CKD stage and age in that 72.3 percent of participants age 60 and older in Stages 4–5 have cardiovas-cular disease.

figure 1.10 Loss of GFR is highly associated with cardiovascular risk factors as demon-strated by a stepwise decline in the preva-lence of diabetes, hypertension, cardiovas-cular disease, and obesity with increasing levels of eGFR.

figure 1.11  Abnormal urinary albumin excretin is strongly associated with other cardiovas-cular risk factors such that a stepwise increase in the prevalence of diabetes, hypertension, cardiovascular disease, and obesity is evident with increasing levels of urinary albumin/creatinine ratio.

table 1.e & figure 1.12 Knowing the relationship between stage of CKD and the likelihood of developing treatable metabolic abnormali-ties is clinically relevant. This table shows the association between stage of CKD and the prevalence of several metabolic disor-ders. Data show there is a discernible rise in the prevalence of hyperkalemia in CKD Stage 2; for the other metabolic abnormali-ties, the CKD stages showing a discernible increase in metabolic abnormalities are as follows: Stage 1, parathyroid hormone, sys-tolic blood pressure, diastolic blood pressure, triglycerides, fasting glucose, waist circum-ference, and anemia; Stage 2, potassium and uric acid; Stage 3, iron deficiency; Stages 4–5, bicarbonate.

table 1.f & figure 1.13  We show here a corre-sponding multivariate analysis quantifying the odds of these metabolic abnormalities according to the presence or absence of eGFR <60. In these models, eGFR values <60 were associated with abnormal potas-sium, bicarbonate, uric acid, calcium, phos-phorus, parathyroid hormone, HDL and hemoglobin levels.

table 1.g & figures 1.14 & 1.15 Here we show a com-parison of prevalence, awareness, treatment, and control of hypertension, dyslipidemia, and diabetes by stage of CKD among U.S. adults. While 80 percent of adults in Stage 3–4 have hypertension, only 20 percent of them are adequately managed, in the sense of being aware of the abnormality and on a successful treatment regime. Management of hyperlipidemia is correspondingly poor, with 18 percent of CKD patients on a suc-cessful treatment regime. Among diabetics, achievement of target glycosylated hemoglo-bin levels is as poor as in the general popula-tion. Overall, there is considerable room for improvement in the management of treatable cardiovascular risk factors in members of the general population with CKD.

Analysis definitions
1.  Hypertension defined as blood pressure ≥130/≥80 for those with CKD and diabetes; otherwise ≥140/≥90, or self-reported treatment for hypertension.
2.  Awareness and treatment are self-reported. Control defined as <130/<80 for those with CKD and diabetes; otherwise <140/<90.
3.  Hyperlipidemia based on elevated LDL (following ATP III guidelines, with CKD considered a CHD risk equivalent), self-reported treatment, or self-reported dieting to lower cholesterol.
4.  Awareness and treatment self-reported. Control defined as meeting the ATP III LDL target: <100 mg/dl (high risk), <130 mg/dl (moderate risk), or <160 mg/dl (low risk).
5.  HDL cholesterol classified according to ATP III guidelines.
6. Total cholesterol classified according to ATP III guidelines.7 Glycohemoglobin classified according to American Diabetes Association guidelines.



figures 1.2–4 NHANES partici-pants age 20 & older. figure 1.5 NHANES 2003–2006 participants age 20 & older. table 1.a NHANES participants 1999–2006. Method One: standard method, using eGFR & microalbuminuria to define CKD. Method Two: gender-specific microalbuminuria used to define CKD. Method Three: same as Method One, but excluding pregnant women. Method Four: same as Method Two, but exclud-ing pregnant women. *Estimate is not reliable.

tables 1.b–c NHANES 1999–2006, participants age 20 & older. PPV: positive predictive value. NPV: negative predictive value. figure 1.6 NHANES 1988–2006, participants age 20 & older.

table 1.d & figures 1.8–11 NHANES 1999–2006, participants age 20 & older. *Estimate is not reliable.

table 1.e & figure 1.12 NHANES participants 1999–2006, age 20 & older. Levels based on criteria proposed by the National Cholesterol Education Program (NCEP) Adult Treatment Panel III (ATP III). Reduced HDL: men <40 mg/d, women <50 mg/d; elevated triglycerides: ≥150 mg/dl; elevated blood pressure: ≥130/85 mmHg; elevated fasting glucose: ≥100 mg/d; elevated waist circumference: men, ≥40 inches (102 cm), women ≥35 inches (88 cm); WHO ane-mia: hemoglobin <13 g/dl for males, <12 g/dl for females. *Estimate is not reliable. table 1.f & figure 1.13 NHANES participants 1999–2006, age 20 & older. Adjusted for age, gender, race, ethnicity, birth outside the U.S., self-reported diabetes, & self-reported hypertension.

table 1.g & figures 1.14–15 NHANES 1999–2006, participants, age 20 & older