Taipei Municipal Zhongxiao Hospital did not have an institutional ethics committee in 2003. The merger of ten municipal hospitals into one unit has made the Taipei City Hospital the largest healthcare organization in Northern Taiwan in January 1, 2005. The first version of institutional review board standard operating procedure of Taipei City hospital was set in August 30, 2006. Taipei City hospital became a teaching hospital of National Yang Ming University in 2006. Thus, these explained why these committees not consulted to review and approve this study. Our study is a prospective study, our dieticians witnessed verbal consent of evaluation of nutritional status. The department of teaching and research, the counterpart of ethical committee, of Taipei Municipal Zhongxiao Hospital approved this research study in 2003. Because our study was a non-invasive one and it only involved measurements of anthropometry and single bioelectric impedance analysis, consent was verbal. The department of teaching and research approved the whole study, including this consent procedure. The fasted blood sample and analyses collected for routine medical purposes. All of the data were anonymized/de-identified prior to access and analysis. Any of the authors did not have access to identifying patient information.
Calculation of sample size required for linear multiple regression models from squared correlation coefficients
A previous study provided correlation coefficients . Triceps skinfold thickness (R=0.84; P< 0.001) significantly correlated with percent fat mass; Mid-arm muscle circumference significantly associated with lean body mass (R=0.79; P< 0.001). The maximum number of predictors in the linear regression models of table 3 and table 4 is five. We chose a desirable significant level of 0.05. Using these data and G*Power 22.214.171.124, the minimum sample size required for building a linear regression model was about 30.
The study subjects were 32 chronic renal failure patients receiving regular hemodialysis in the renal unit of Taipei Municipal Zhongxiao Hospital from July 2003 to August 2004. Eighteen diabetic patients were recruited in this study. The study subjects were selected because they agreed to receive the measurements. Our dieticians witnessed oral consent for the measurements. All subjects were without active medical or surgical illness, and none were receiving corticosteroids that could result in body mass depletion.
Variables of BIA, Upper Arm Anthropometry and Laboratory Indexes
Fasted blood samples were taken and sent to our laboratory for the following analyses: serum albumin, total protein, blood urea nitrogen, creatinine, cholesterol, and triglyceride. Serum albumin concentrations were measured by the bromcresol green (BCG) method.
A well-trained dietician carried out the following measurements on each patient in triplicate after a dialytic session on an outpatient basis: height, weight, mid-arm circumference (MAC), TSFT, percent fat mass, total body water (TBW), and FFM. The dry weight of each patient was individually assessed according to the cardiothoracic ratio, pretibial edema, facial edema, pulmonary edema, jugular vein pressure, and a decrease in blood pressure during the hemodialysis session or inter-dialytic period.
Measurements of percent fat mass, TBW, FFM were obtained with single-frequency BIA (BA-200, Mesmed System Co., Ltd.). The procedure of BA-200 is simple and easy. The BA-200 was done with the patient relaxed in a sitting position with legs separated and arms abducted from the body. Measurements were done on the nonvascular access site. The proximal (detector) electrodes were placed on the leg and forearm. A low- amplitude, single frequency, imperceptible alternating current (800μA at 50 kHz) was introduced via the source electrodes at the wrist and ankle. The method for estimating total body water is a BIA estimate.
Height was measured to an accuracy of 0.1 cm by using a wall-mounted stadiometer (Holstain, Crymych, UK). Weight was recorded using a digital scale accurate to 0.02 kg (FW-122 series, floor scale, maximum measured weight 122 kg). Triceps skin-fold thickness of the arm without a vascular access was measured to the nearest 0.5 mm with a Lange skin-fold Caliper, calibrated to exert a pressure of 10 g/mm2 of jaw surface. A flexible plastic-coated tape was applied to measure MAC of the non-access arm to the nearest 0.1cm with sufficient tension to touch, but not indent the skin. From these measurements, the following indices were calculated: MAMC (cm), mid-arm area (MAA; mm2), mid-arm muscle area ( MAMA, mm2), cMAMA (mm2), mid-arm fat area (MAFA, mm2), BMI (kg/m2), and FFMI (kg/m2) .
Formulae for these are as follows :
MAMC (cm) = MAC (cm) - (π × TSFT (mm) /10)
MAA (mm2) = [MAC (mm)]2 ÷ 4π
MAMA (mm2) = [MAC (mm)- π × TSFT (mm)]2 ÷ 4π
MAFA (mm2) = MAA – MAMA = [MAC (mm)]2 ÷ 4π- [MAC (mm) - π×TSFT (mm)]2 ÷ 4π
Corrected MAMA (males, mm2) = [MAC (mm)- π × TSFT (mm)]2 ÷ 4π- 1000
Corrected MAMA (female, mm2) = [MAC (mm)- π × TSFT (mm)]2 ÷ 4π- 650
Data were analyzed using SPSS 15.0 for Windows. Quantitative results were presented as mean or median. Because of considerable skewness in the data separated by gender, comparisons were made using Mann Whitney U tests. We used Pearson correlation to assess the relationship among derived estimates by BIA and measurements by upper arm anthropometry and clinical parameters. Because BMI includes fat mass and lean mass in its measurement, we used BMI in the linear regression models for comparison. Multiple regressions were used to quantify the relative importance of variables of upper arm anthropometry, biochemical parameters, gender, and BMI in the prediction of body composition obtained from BIA. A value of P < 0.05 was taken to indicate statistical significance. We checked variance inflation factor, condition index and eigen value for collinearity. Mid-arm muscle circumference, MAC, MAMA, cMAMA, and predialysis serum creatinine had collinearity. Mid-arm fat area, MAC and TSFT had collinearity. Height and BMI had collinearity. Thus, we treated collinearity by putting them separately in the model and eliminating the variable, which caused inaccurate computations of the regression coefficients.
Test-Retest Reliability of Total Body Water
Because the same operator obtained the measurement of total body water, we can avoid operator- dependent (inter-observer) random measurement error. For the test-retest reliability trials, the intra-class correlation coefficients (Rho) of repeated total body water with 1-way random model and its 95% confidence interval were 0.997 (0.995-0.998).
The norms or the thresholds of upper arm anthropometry, clinical parameters and BIA derived values
The data set of the Second National Health Survey (1986-1988) of a representative pool of the healthy Taiwanese population provided percentiles for TSFT, MAC and MAMC of Taiwanese . The survey is of apparently 20653 healthy subjects (10398 females, 10255 male) and stratified by different age groups. As a general rule, measurements below the 10th percentile of normal population and below the 50th percentile of MHD patients are considered to represent significant malnutrition. Furthermore, racial differences in body composition and body proportions also exist. A previous study pointed out upper arm anthropometric standards established for the United States population of whites . Another publication provided age-, gender-, diabetics-, race- and dialysis treatment modality- specific reference norms for TSFT and MAC of the stable dialysis population . Patients with percent fat mass smaller than 10 % were defined as energy malnutrition . Based on WHO definition, overweight was defined as a BMI more than 23 kg/m2, obesity as a BMI more than 25 kg/m2, and underweight as a BMI below 18.5 kg/m2 in adult Asians. However, evaluation of prevalence of overweight, obesity or underweight, based on WHO definition for a general population, may be not proper for HD population.
Based on the data from healthy population of whites, FFMI (kg/m2) was “low” at <16.7 (men) and < 14.6 (women);”normal” between 16.7 and 19.8 (men) and 14.6 and 16.8 (women); and “high” at >19.8 (men) and > 16.8 (women) . Ten Percentile values of FFMI (kg/m2) for the healthy population of whites by different ages categories were as follows : between age 18 –34 were 17.2(men) and 14.1(women); between age 35-54 were 17.6 (men) and 14.7 (women); between age 55-74 were 17.6(men) and 14.6 (women); above age 75 were 16.9 (men) and 13.7(women). Pre-dialysis serum albumin concentrations < 3.8 g/dl reflect visceral protein malnutrition .