2010 USRDS Annual Data Report
This Chapter
Download
Download PDF
Download all slides
*corresponding data in Excel included
Search This Page
Search All
Translate

Figure List
Figure 1.1 Distribution of NHANES participants with diabetes, congestive heart failure, & markers of CKD, with GFR estimated by MDRD, CKD-EPI, & cystatin C equation
Table 1.a Prevalence & odds of CKD in NHANES 19992006 participants, by method used to estimate GFR, CKD stage, age, gender, race/ethnicity, & risk factor (percent of participants)
Table 1.b Sensitivity & specificity of different population characteristics for identifying eGFR < 60 ml/min/1.73 m2 & urinary ACR =30 mg/g
Figure 1.2 Cumulative distribution curves of NHANES 19881994 participants, by eGFR & method used to estimate GFR
Figure 1.3 Cumulative distribution curves of NHANES 19992002 participants, by eGFR & method used to estimate GFR
Figure 1.4 Cumulative distribution curves of NHANES 20032006 participants, by eGFR & method used to estimate GFR
Figure 1.c Prevalence (%) of diabetes, hypertension, & cardiovascular disease in NHANES 19992006 participants, by age, gender, race/ethnicity, CKD status, & method used to estimate GFR
Figure 1.5 Prevalence of comorbidity in NHANES 19992006 participants, by risk factor, eGFR, & method used to estimate GF
Figure 1.6 Prevalence of comorbidities in NHANES 19992006 participants, by risk factor, expanded eGFR categories, & method used to estimate GFR
Figure 1.7 Distribution of NHANES 19992006 participants, by risk factor, eGFR, & method used to estimate GFR
Figure 1.8 Prevalence of comorbidity in NHANES 19992006 participants, by risk factor & albumin/creatinine ratio
Figure 1.d Abnormalities of selected clinical & biochemical parameters in NHANES participants, by eGFR & method used to estimate GFR (percent of participants)
Figure 1.9 Abnormalities of selected clinical & biochemical parameters in NHANES 19992006 participants, by expanded eGFR categories & method used to estimate GFR
Figure 1.e Awareness, treatment, & control of hypertension, hypercholesterolemia, HDL cholesterol, & diabetes, by CKD stage & method used to estimate GFR (percent of NHANES participants)
Figure 1.10 Odds of awareness, treatment, & control of hypertension, by CKD stage & method used to estimate GFR
Figure 1.11 Odds of awareness, treatment, & control of hypercholesterolemia, by CKD stage & method used to estimate GFR
Figure 1.12 Odds of meeting target HDL & total cholesterol levels, by CKD stage & method used to estimate GFRs
Figure 1.13 Odds of diabetes control, by CKD stage & method used to estimate GFR
Figure 1.14 Predicting death: sensitivity & specificity of different eGFR thresholds: MDRD equation
Figure 1.15 Predicting death: sensitivity & specificity of different eGFR thresholds: CKD-EPI equation
Figure 1.16 Predicting death: sensitivity & specificity of different cystatin C thresholds
Figure 1.17 Predicting death: sensitivity & specificity of different ACR thresholds
Figure 1.18 Mortality rates in NHANES 19881994 participants, by age, gender, race, & eGFR: MDRD equation
Figure 1.19 Mortality rates in NHANES 19881994 participants, by age, gender, race, & eGFR: CKD-EPI equation
Figure 1.20 Mortality rates in NHANES 19881994 participants, by age, gender, race, & eGFR: cystatin C
Figure 1.21 Mortality rates in NHANES 19881994 participants, by age, gender, race, & albumin/creatinine ratio

Chapter One
Chronic kidney disease in the general population  Top

Sections this chapter: 

In this chapter we use data from the National Health and Nutrition Examination Survey (NHANES), a valuable source of information for assessing disease burden and high-risk subsets among representative U.S. adults. On the next page we begin by showing the overall burden and interactions of diabetes, congestive heart failure, and CKD three interrelated chronic diseases. Data vary only slightly based on the estimating equation used to define CKD: the older MDRD Study equation, which underestimates glomerular filtration rates (eGFRs) over 60 ml/min/1.73 m2, compared to the newer CKD-EPI equation, which partially addresses this issue.

The prevalence of diabetes in the U.S. continues to grow, reaching 7.7 percent of the general population in 20032006. The prevalence of congestive heart failure has also increased, to 2.5 percent.

Estimating the CKD population is more challenging. While a decline in kidney function is common in the elderly, this population also has the greatest burden of chronic disease, confounding the question of whether CKD is simply part of aging or a true chronic disease coexisting with others such as diabetes and hypertension. In any case, its implications are considerable.

Compared to the MDRD equation, the new CKD-EPI equation for estimating GFR identifies, in populations younger than 60, a greater number of patients with GFRs above 60 ml/min/1.73 m2, and has greater sensitivity to detect CKD, though similar specificity. Use of CKD-EPI thus shows a greater burden of comorbidity among patients with eGFRs less than 60 ml/min/1.73 m2, as the placement of healthier patients into the above-60 group leaves behind a denominator of patients with more chronic disease. For the same reason, use of the CKD-EPI equation also shows a greater burden of biochemical abnormalities among those patients still classified as having lower eGFRs.

With potent mortality classification systems, "normal" values should accurately predict survival (high specificity) and "abnormal" values should accurately predict death (high sensitivity); in public health settings, classification based on a threshold with the highest combined sensitivity and specificity is a logical choice. We evaluate sensitivity and specificity of different eGFR and ACR thresholds for discriminating death or survival. For the cystatin C-based eGFR, a threshold value of 90 ml/min/1.73 m2 matches to a threshold of 85 with the creatinine-based MDRD equation, a value of 91 with the CKD-EPI equation, and a urinary albumin-creatinine ratio of 8 mg/g.

These data demonstrate that the equation used to estimate GFR has an important impact on the prediction of death. The CKD-EPI equation provides a more consistent pattern of mortality rates than does the MDRD equation. As it addresses the issue of falsely low eGFR measurements (particularly in younger populations), and improves the predictive value of eGFRs, the CKD-EPI equation should be considered to replace the older MDRD equation. Additionally, socioeconomic factors should be addressed in future assessments of CKD and its predictive power related to death and other outcomes.

Exploring the implications of CKD, diabetes, and cardiovascular disease in the general population, this chapter sets the stage for Chapter Two, in which we discuss the implications of CKD in datasets that are less well defined in terms of biochemical data, but that provide extensive information on morbidity, interventions, and costs not contained in the NHANES data or other samples.

Figure 1.1 Distribution of NHANES participants with diabetes, congestive heart failure, & markers of CKD, with GFR estimated by MDRD, CKD-EPI, & cystatin C equations. See page 166 for analytical methods. NHANES participants age 20 & older.

Table 1.a Prevalence & odds of CKD in NHANES 19992006 participants, by method used to estimate GFR, CKD stage, age, gender, race/ethnicity, & risk factor (percent of participants)

Data on the different stages of CKD among NHANES participants show that, using a single creatinine-based eGFR and the MDRD equation, 3.2, 4.1, 7.8, and 0.5 percent, respectively, have CKD of Stages 1, 2, 3, and 45. With the CKD-EPI equation, the corresponding proportions are 4.3, 3.2, 6.3, and 0.6 percent, indicating that CKD-EPI tends to give a higher eGFR estimate than the MDRD equation. When using a single eGFR based on cystatin C, the corresponding proportions are 4.0, 3.9, 6.4, and 0.6 percent. Multivariate associations of CKD for all three eGFR methods include older age, female gender, self-reported diabetes, hypertension, and cardiovascular disease.

Table 1.b Sensitivity & specificity of different population characteristics for identifying eGFR < 60 ml/min/1.73 m2 & urinary ACR =30 mg/g

Table 1.b shows the sensitivity and specificity of different screening strategies to identify eGFR values less then 60 ml/min/1.73 m2 or albumin creatinine ratio (ACR) values of 30 mg/g and above. In general, the screening strategies show greater sensitivity for detecting eGFRs less than 60 ml/min/1.73 m2 with the CKD-EPI equation than with the MDRD or cystatin C methods, while specificity values tend to be similar across all three methods.

Tables 1.ab; see page 166 for analytical methods. NHANES 19992006 participants age 20 & older. *Estimate not reliable.

Figure 1.2 Cumulative distribution curves of NHANES 19881994 participants, by eGFR & method used to estimate GFR

Figure 1.3 Cumulative distribution curves of NHANES 19992002 participants, by eGFR & method used to estimate GFR

Figure 1.4 Cumulative distribution curves of NHANES 20032006 participants, by eGFR & method used to estimate GFR

Figures 1.24; see page 166 for analytical methods. NHANES participants age 20 & older.

In cumulative frequency distributions of eGFR in U.S. adults, creatinine-based CKD-EPI and cystatin C methodologies for eGFR calculation yield higher estimates of GFR than those achieved when using the creatinine-based MDRD method.

Comorbidity burden  Top

Table 1.c Prevalence (%) of diabetes, hypertension, & cardiovascular disease in NHANES 19992006 participants, by age, gender, race/ethnicity, CKD status, & method used to estimate GFR.

Diabetes, hypertension, and cardiovascular disease are much more common in persons with CKD than in those without. In general, there is a trend towards higher prevalence estimates with rising CKD stage; this is particularly true in patients with hypertension and cardiovascular disease. Table 1.c; see page 166 for analytical methods. NHANES 19992006 participants age 20 & older. *Estimate not reliable.

Figure 1.5 Prevalence of comorbidity in NHANES 19992006 participants, by risk factor, eGFR, & method used to estimate GFR
With both creatinine-based MDRD and CKD-EPI estimates, the prevalence of diabetes, hypertension, or cardiovascular disease is noticeably higher in subjects with eGFRs below 60 ml/min/1.73 m2 than among those whose eGFR equals or exceeds 60. Approximately 6166 percent of NHANES participants with an eGFR less than 60, for example, have hypertension, compared to 25 percent in those with an eGFR of 60 or greater, and the prevalence of cardiovascular disease is more than five times greater in those with advanced CKD, at 3338 versus 67 percent.

Figure 1.6 Prevalence of comorbidities in NHANES 19992006 participants, by risk factor, expanded eGFR categories, & method used to estimate GFR

The prevalence of disease rises with CKD severity. In participants with eGFRs less than 30, 30 < 45, and 45 < 60, for example, 37, 24, and 17 percent report having diabetes, compared to 6.1 percent of those with an eGFR of 60 or above. And in participants with an eGFR less than 30 ml/min/1.73 m2, approximately 83 percent have hypertension and 63 percent have cardiovascular disease, compared to 25 and 7.0 percent, respectively, of participants with an eGFR of 60 or greater.

Figure 1.7 Distribution of NHANES 19992006 participants, by risk factor, eGFR, & method used to estimate GFR

More than three of four NHANES participants with an eGFR less than 60 ml/min/1.73 m2 are age 60 or older, while only 3.0 percent are age 39 or younger.

Figure 1.8 Prevalence of comorbidity in NHANES 19992006 participants, by risk factor amp; albumin/creatinine ratio

Figures 1.58; see page 166 for analytical methods. NHANES 19992006 participants age 20 & older

The prevalence of comorbid illness among NHANES participants rises with albumin/creatinine ratio (ACR). Four percent of those with an ACR less than 10 mg/g have diabetes; this rises to 24 percent of those whose ACR is 30 or greater. Hypertension and cardiovascular disease are evident in 23 and 6.1 percent of individuals with an ACR below 10, compared to 49 and 21 percent in those with an ACR of 30 or greater.

Clinical & biochemical abnormalities  Top

Table 1.d Abnormalities of selected clinical & biochemical parameters in NHANES participants, by eGFR & method used to estimate GFR (percent of participants). See page 166 for analytical methods. NHANEs 19992006 participants age 20 & older; for cystatin C, NHANES 19992002 participants. *Estimate not reliable; **significant (p<0.01)

Regarding the prevalence of clinical and laboratory abnormalities in subjects with and without an eGFR less than 60 ml/min/1.73 m2, there are associations with the following abnormalities: high potassium, uric acid, and phosphorus, reduced HDL, and low hemoglobin levels.

Figure 1.9 Abnormalities of selected clinical & biochemical parameters in NHANES 19992006 participants, by expanded eGFR categories & method used to estimate GFR. See page 166 for analytical methods. NHANEs 19992006 participants age 20 & older; for cystatin C, NHANES 19992002 participants. *Estimate not reliable.

Regarding associations between selected clinical and biochemical abnormalities, the proportions of patients with each abnormality are broadly independent of the method used to estimate GFR. Conditions associated with low eGFR include: elevated systolic blood pressure, reduced HDL, elevated triglycerides, elevated fasting glucose, and elevated waist circumference.

Awareness, treatment, & control of disease conditions  Top

Table 1.e Awareness, treatment, & control of hypertension, hypercholesterolemia, HDL cholesterol, & diabetes, by CKD stage & method used to estimate GFR (percent of NHANES participants)

Here we use NHANES data from 19992006 to evaluate awareness, treatment, and control of disease conditions, using CKD stages defined with two creatinine-based methodologies to estimate GFR. With the MDRD method, 80.5 percent of participants with CKD of Stages 34 have hypertension; only 20 percent, however, are aware of their condition and on a successful treatment regime. With the CKD-EPI method, 84.4 percent of Stage 34 participants have hypertension, while 19.5 percent are aware of their condition and receiving adequate treatment. Among patients with earlier stages of CKD, both MDRD and CKD-EPI show that 64 percent have hypertension, more than one-third are unaware of their condition, 14 percent are not treated, and 11 percent are on a successful treatment regime.

With both MDRD and CKD-EPI, 80 percent of participants with Stage 34 CKD have hypercholesterolemia (based on elevated LDL), but only 1822 percent are treated and brought to levels recommended by clinical practice guidelines. In those with less severe CKD, 4849 percent have hypercholesterolemia, while less than one-third are aware of their condition and adequately controlled. Approximately 20 percent of CKD patients have high (240+) total cholesterol levels and HDL cholesterol below the recommended levels, while 63 percent of participants with Stage 12 CKD and 47 percent of those with Stage 34 CKD have glycohemoglobin levels above the recommended 7 percent guideline.

Table 1.e; see page 166 for analytical methods. NHANES 19992006 participants age 20 & older; those with Stage 5 CKD excluded.

Figure 1.10 Odds of awareness, treatment, & control of hypertension, by CKD stage & method used to estimate GFR

Regardless of the method used to estimate GFR, participants with CKD are 45 times more likely to have hypertension than those without. Patients with Stage 12 CKD are more likely to be aware of their condition and on a successful treatment protocol compared to those with Stage 34 CKD. And those with Stage 12 CKD are also nearly twice as likely to receive anti-hypertensive medications in the form of an ACE inhibitor, angiotensin receptor blocker, orrenin inhibitor than those with more advanced CKD.

Figure 1.11 Odds of awareness, treatment, & control of hypercholesterolemia, by CKD stage & method used to estimate GFR

With the MDRD equation, the likelihood of hypercholesterolemia is more than four times greater in participants with Stage 34 CKD than in those with no CKD, and it is three times greater with CKD-EPI. Both methods show that participants with Stage 34 CKD are more likely to be aware of and suitably treated for their condition than those with less severe CKD, but the likelihood of receiving a lipid lowering medication is similar for all CKD stages.

Figure 1.12 Odds of meeting target HDL & total cholesterol levels, by CKD stage & method used to estimate GFR

With both methods, the odds of meeting a high density lipid (HDL) level of 40 mg/dl or higher are greater in participants with Stage 34 CKD than in those with less severe CKD; the opposite is true for total cholesterol levels. Participants with Stage 12 CKD are more likely to have a total cholesterol of less than 240 mg/dl than those with CKD of Stages 34.

Figure 1.13 Odds of diabetes control, by CKD stage & method used to estimate GFR

And with both eGFR methods, diabetes control is nearly twice as likely in participants with less severe CKD.

Figures 1.1013; see page 166 for analytical methods. NHANES 19992006 participants age 20 & older; those with Stage 5 CKD excluded. For Figures 1.101.11, *participants with hypertension; **participants with hypercholesterolemia.

Predictive models for CKD; mortality  Top

Figure 1.14 Predicting death: sensitivity & specificity of different eGFR thresholds: MDRD equation

Figure 1.15 Predicting death: sensitivity & specificity of different eGFR thresholds: CKD-EPI equation

Figure 1.16 Predicting death: sensitivity & specificity of different cystatin C thresholds

Figure 1.17 Predicting death: sensitivity & specificity of different ACR thresholds

Figures 1.1417; see page 166 for analytical methods. NHANES III (19881994) participants age 20 & older.

For screening purpose, it can be useful to know the efficacy of different threshold levels for predicting death or survival. For death within a finite time interval, a threshold where individuals classified as "normal" show low mortality rates (a high proportion of true negatives, high specificity for predicting death) and those classified as "abnormal" show high mortality rates (a high proportion of true positives, high sensitivity for predicting death) might be attractive for defining subgroups in which intensive follow-up and treatment may be appropriate, as well as for classification purposes.

Figures here show the sensitivity and specificity, for predicting death, of different threshold levels of commonly-used renal function measures among representative U.S. adults between 1988 and 1994, followed through 2006. Considered as continuous variables, the C-statistic for death (an index of discrimination between survival and death with perfection denoted by a value of 1) was 0.81 with the CKD-EPI equation, 0.79 with serum cystatin C, 0.75 with the MDRD equation, and 0.69 with ACR. For eGFR and ACR, respective thresholds of 60 ml/min/1.73 m2 and 30 mg/g are highly specific but highly insensitive, as the vast majority of subjects who died during the observation period had an eGFR greater than 60 and an ACR less than 30. Recognizing that gains in sensitivity are always accompanied by losses in specificity, and vice versa, thresholds of maximizing sensitivity plus specificity are approximately 90 ml/min/1.73 m2 for eGFR, and 10 mg/g for ACR. For cystatin C, the threshold value showing the highest maximum combined value of sensitivity and specificity for predicting death is approximately 0.9 mg/l.

Figure 1.18 Mortality rates in NHANES 19881994 participants, by age, gender, race, & eGFR: MDRD equation

Here we show proportions of NHANES 19881994 participants who died during a follow-up interval extending to December 31, 2006, and categorize them by interactions of race, gender, and renal function. In general, because of low mortality rates, caution must be used when interpreting mortality in subgroups younger than 40.Restricting comparisons to participants age 40 or older, and using the MDRD equation to estimate GFR, the expected monotonic association between lower eGFR category and higher mortality is present only for white females age 4059 and non-white females age 40 and older; for all other categories, mortality is highest with an eGFR less than 60 ml/min/1.73 m2, lowest with an eGFR 6089, and intermediate with an eGFR of 90 or above.

Figure 1.19 Mortality rates in NHANES 19881994 participants, by age, gender, race, & eGFR: CKD-EPI equation

Similarly restricting comparisons to those age 40 or greater and using the CKD-EPI equation to estimate GFR, the expected monotonic association between lower eGFR and higher mortality is absent in males age 4059 (with higher mortality in the subgroup with an eGFR of 90 or greater than among those with an eGFR of 6089).

Figure 1.20 Mortality rates in NHANES 19881994 participants, by age, gender, race, & eGFR: cystatin C Using a similar strategy with the cystatin C-based eGFR, the expected monotonic association pattern is present in each subgroup except white and non-white males age 4059.

Figure 1.21 Mortality rates in NHANES 19881994 participants, by age, gender, race, & albumin/creatinine ratio

Figures 1.1821; see page 166 for analytical methods. NHANES III (19881994) participants age 20 & older, followed to December 31, 2006.

When spot urinary albumin/creatinine ratio (ACR) is categorized as less than 30, 30299, and 300 mg/g or more, and comparisons are restricted to those age 40 or older, the expected monotonic association between rising ACR and rising mortality is present in each of the subgroups studied.