Study participants: Third National Health and Nutrition Examination Survey
The National Center for Health Statistics of the United States Centers for Disease Control and Prevention conducted the Third National Health and Nutrition Examination Survey (NHANES III) between 1988 and 1994 as a cross-sectional study [25
]. Its study population consisted of a representative sample of the general non-institutionalized civilian US population, using a stratified multistage probability design [25
]. Very young, elderly, non-Hispanic black and Mexican American persons were deliberately oversampled for improved estimate precision in these groups. More detailed information about the study participants and methods has been published before [26
The NHANES III analyses in this report were limited to 17 030 adult participants ≥20 years who were examined at a mobile examination center as part of the regular study protocol. Since only the first enrollment phase of NHANES III (1988–91) included apolipoprotein measurements, all examinees enrolled in the second phase were excluded. Of the remaining 8213 participants, those with missing serum apolipoprotein A1 (n = 609) and B (n = 28) as well as serum creatinine (n = 185) measurements were sequentially excluded. The final study population consisted of 7391 participants (90% of Phase I participants were aged over 20 years). Of the 7391 participants, 7023 had information on all covariates available and were included in the multivariable-adjusted analyses unless otherwise indicated.
Study participants: Atherosclerosis Risk in Communities study
The Atherosclerosis Risk in Communities (ARIC) study consists of population-based samples from four US communities. The detailed design of ARIC has been described [27
]. In summary, 15 792 women and men were recruited between 1987 and 1989 based on probability sampling techniques. Participants underwent standardized examinations in field center clinics, including measurements of demographic, lifestyle and physical factors as well as laboratory parameters. They were re-examined at follow-up visits every 3 years. This report is based on the fourth examination of the cohort (1996 through 1998), when apolipoprotein A1 and B as well as UACR were measured. The overall ARIC cohort follow-up rate was 80% for Visit 4 [28
]. All participants with missing serum creatinine and apolipoprotein measurements were excluded. Of the remaining individuals, information on all relevant covariates was available for 10 292 participants.
In 2005/06, the ARIC Carotid MRI follow-up study examined 2066 ARIC participants, aged 60–84 years. Participants were selected using a stratified plan to oversample for plaque based on carotid thickness at a prior ultrasound examination (1993–98) [29
]. Of these participants, 1659 had available serum creatinine and relevant covariate measurements and were examined in the prospective analyses.
Information was obtained on a wide variety of sociodemographic characteristics (such as age, race–ethnic group, gender, medication use) by administration of a standard questionnaire in NHANES III and ARIC participants. Race–ethnicity was reported in the categories of non-Hispanic white, non-Hispanic black, Mexican American and others for NHANES III. Race was reported in the categories of white and black for ARIC. The interview was followed by a detailed physical examination including blood specimen collection in both NHANES III and ARIC [27
In NHANES III, diabetes mellitus was defined by a serum glycated hemoglobin >6.5% according to the 2010 guidelines of the American Diabetes Association [30
] and/or self-reported use of anti-diabetic medications in conjunction with a known physician diagnosis of diabetes. In ARIC, diabetes was defined as fasting serum glucose ≥6.99 mmol/L a self-reported previous diagnosis of diabetes or use of anti-diabetic medications.
Mean blood pressure values were obtained from up to six measurements during the home interview and during the subsequent evaluation at the mobile examination center in NHANES III. Hypertension was defined by a mean systolic blood pressure >140 mmHg, mean diastolic blood pressure >90 mmHg and/or by self-reported use of anti-hypertensive medications in conjunction with a known physician diagnosis of hypertension for NHANES III. A similar definition was used for ARIC based on the mean of two blood pressure measurements.
For the NHANES III analysis, current alcohol use was defined as self-reported consumption of 12 alcoholic beverages or more during the previous 12 months. For ARIC, current alcohol intake was defined as self-reported current consumption of alcohol. Current smoking was defined as self-reported current smoking of any cigarettes for NHANES III and ARIC. Body weight and height were measured as defined by a standard protocol, and the body mass index (BMI, kg/m2) was then calculated in both NHANES III and ARIC. Obesity was defined as BMI ≥30 kg/m2.
Serum apolipoprotein A1 and B levels were measured regardless of the examinee's fasting status in NHANES III by three different methods: radial immunodiffusion (RID), rate immunonephelometry (INA) and the World Health Organization–International Federation of Clinical Chemistry (WHO–IFCC) method. Results using the RID and INA methods were adjusted to the WHO–IFCC method [31
]. The reported coefficients of variation ranged from 2.7 to 20.8% (median <5.2%) [31
]. Plasma apolipoprotein A1 and B were measured in ARIC by an immunonephelometric assay with a BNII nephelometer (Siemens Healthcare Diagnostics, Deerfield, IL). Apolipoprotein measurements were begun in September 2009 and completed by January 2010. The reliability coefficients for 397 blinded replicates were 0.88 (0.92 after removing 8 pairs of outliers) for apolipoprotein A1 and 0.89 (0.96 after removing 11 pairs of outliers) for apolipoprotein B. This immunonephelometric methodology is in contrast to radioimmunoassay and immunoturbidometry methods used for previous ARIC apolipoprotein measurements among Visit 1 and 2 participants, techniques that demonstrated lower precision than current standardized assays.
Serum triglyceride, HDL and total cholesterol levels were measured regardless of the examinee's fasting status [31
]. In ARIC, >95% of participants were fasting at least 8 h prior to the blood draw for the plasma lipid measurements. Due to their skewed distribution, serum triglycerides were log transformed for all analyses in both ARIC and NHANES III. Serum LDL was calculated by using the Friedewald equation [31
]. Serum LDL was only calculated if triglycerides were <4.52 mmol/L and if the participants were fasting for >9 h prior to blood draw for NHANES III [31
Urinary albumin and creatinine were measured as described before [33
]. The UACR was calculated and microalbuminuria (MA) was then defined as UACR ≥30 mg albumin/g creatinine.
The modified kinetic Jaffe reaction using a Hitachi 737 analyzer (Boehringer Mannheim Corp., Indianapolis, IN) was employed to measure serum creatinine concentrations in NHANES III [31
]. In ARIC, blood creatinine was measured with the modified kinetic Jaffe reaction from plasma samples in Visit 4 using a DACOS Chemistry Analyzer (Dart Reagent Systems; Coulter Electronics, Hialeah, FL) [34
]. For the ARIC Carotid MRI visit (MRI visit), blood creatinine was measured with the modified kinetic Jaffe reaction using the Olympus AU400e automated chemistry analyzer.
Standard creatinine was calculated for NHANES III using the equation: standard creatinine = 0.96 × serum creatinine − 0.184 [37
]. ARIC Visit 4 creatinine measurements were calibrated for interlaboratory differences to the Cleveland Clinic formula by subtraction of 21.22 μmol/L [38
] and standardized to the Roche enzymatic method by multiplication with 0.95 [37
]. The MRI visit creatinine measurements did not require any standardization.
In the cross-sectional analyses of NHANES III and ARIC Visit 4, the following outcome measures were evaluated: continuous eGFR, continuous UACR and the presence of CKD or MA. Moreover, incident CKD and MA were evaluated between ARIC Visit 4 and the MRI visit (average follow-up time 8.6 years).
The eGFR was calculated using the CKD-EPI equation: eGFR = 141 × min(Scr/κ
× 1.018 [if female] × 1.159 [if Black], where Scr is serum creatinine, κ
is 0.7 for females and 0.9 for males, α is −0.329 for females and −0.411 for males, min indicates the minimum of Scr/κ
or 1 and max indicates the maximum of Scr/κ
or 1 [2
Persons with an eGFR of <15 mL/min/1.73m2 were treated as if their eGFR was 15 mL/min/1.73m2 (n = 13 in NHANES III, n = 19 in ARIC and n = 3 at the MRI visit). Similarly, participants with an eGFR of >200 mL/min/1.73m2 were treated as if their eGFR was 200 mL/min/1.73m2 (n = 1 in NHANES III, n = 0 in ARIC and n = 0 at the MRI visit).
CKD was defined as an eGFR of <60 mL/min/1.73m2
according to the most recent Kidney Disease Outcome Quality Initiative guidelines [39
]. The definition of the term ‘CKD’ used in our article corresponds to CKD Stages 3–5. Incident CKD was defined as eGFR below the threshold of 60 mL/min/1.73m2
at the MRI visit in persons free of CKD at ARIC Visit 4.
All NHANES III analyses were performed using appropriate sampling weights to account for the complex sampling design according to the NHANES III analytic and reporting guidelines [25
]. The application of sampling weights was not necessary for the cross-sectional ARIC Visit 4 analyses. For the longitudinal ARIC Carotid MRI follow-up study, the appropriate sampling weights were applied to account for the sampling design [40
Serum apolipoprotein levels were evaluated in quartiles. Within each apolipoprotein quartile, the study sample was characterized by determining age-standardized mean values or proportions of the outcomes and all covariates. For NHANES III, age standardization was based on the 1980 Census population, as specified in the NHANES III analytic and reporting guidelines [25
For the evaluation of trends across apolipoprotein quartiles, P-values were calculated by adjusted multiple linear and logistic regression: adjustment variables are detailed in the table footnotes.
Associations between apolipoprotein quartiles and eGFR or proportion of participants with CKD were first obtained from adjusted multiple linear and logistic regression analyses including age, gender and race as covariates. Subsequently, multivariable-adjusted analyses were conducted including additional covariates thought to be risk factors for CKD (length of education, obesity, diabetes, hypertension, smoking, alcohol, serum triglycerides) in the full model. P-values for trends of associations across quartiles of the exposure variables were calculated from regression models including a quartile indicator variable and adjusted for the variables listed in the table footnotes.
Collinearity was determined for triglycerides, LDL, HDL and total cholesterol as additional covariates using variance inflation factors. Multicollinearities with variance inflation factors exceeding 10 prevented the inclusion of LDL, HDL and total cholesterol into the full model.
Interactions were tested across strata of race, obesity, diabetes, hypertension, age and gender by including a multiplicative interaction term into the age-, sex- and race-adjusted regression models.
Bootstrap analysis was performed to compare effect size coefficients of the fourth quartile between apolipoprotein and traditional lipids applying 50 replications in ARIC Visit 4 and between 45 and 50 replications in NHANES III analyses.
Reported P-values are two sided. Statistical analyses were performed using STATA version 11.1, Special Edition (StataCorp LP, College Station, TX).