2023 Volume 46 Issue 11 Pages 1609-1618
The modified Cockcroft–Gault (CG) equation, previously developed for an aged-oriented cohort, was validated in a newly obtained dataset. Estimates of creatinine clearance (CCr) using this equation were found to be more accurate than those determined using the conventional CG equation, particularly for patients exceeding 65 years of age. We identified a subset of patients in this cohort whose estimates were inadequate. Using statistical analysis, we found that the deviation from estimates was attributed to a decreased albumin level. In addition, we determined a reduced albumin cutoff value for the modified CG equation to obtain a good estimate. Univariate linear regression analysis was applied to measure the CCr in this cohort and identify parameters related to body composition, and we found that extracellular water (ECW)/total body water (TBW) and body fat (%) were relevant. Using measured values of ECW/TBW and body fat (%), a multivariate linear regression (MLR) estimating equation was developed based on the modified CG equation. This equation was applied to a cohort over 65 years of age, and it was found that a good estimate was obtained for older patients with low albumin levels. Thus, we propose a flow diagram that illustrates conditions for selecting an appropriate estimating equation from among the CG, modified CG, and MLR equations.
Given that renal function, particularly in older individuals, tends to decrease, even in the absence of renal diseases, there is considerable concern regarding the effects of senescence on changes in the pharmacokinetics and pharmacodynamics (PK/PD) of drugs that undergo renal excretion.1,2) While an estimate of the glomerular filtration rate (GFR) is used to assess renal function, the Cockcroft–Gault (CG) equation3) for predicting creatinine clearance (CCr) from serum creatinine (SCr), with variables of weight, age, and sex, was proposed in 1976 and has been applied in clinical practice. However, poor estimates have been obtained, particularly for older individuals.4,5) This discrepancy could be attributed to changes in body composition and its influence on body weight and SCr, known variables in the CG equation.6)
Physiological functions generally decline during adulthood. The onset of sarcopenia, characterized by the loss of muscle mass in patients with diabetes, has been associated with a decrease in markers of renal function, such as urinary albumin level, urinary protein level, and estimated GFR (eGFR).7) SCr and albumin, which are included as variables in the renal function-estimating equations, are predicted to be affected by body composition, such as muscle mass,5,8) and thus renal function estimates are likely to result in discrepancies in patients with substantial changes in body composition. Therefore, incorporating parameters concerning body composition into estimating equations could improve the estimation.
Studies on the relationship between renal function and body composition have been performed from various perspectives, such as the correlation between GFR and lean fat mass,9,10) renal function prediction based on body cell mass11) and the relationship between body fat (%), and the Modification of Diet in Renal Disease (MDRD) Study equation.12) However, these relationships were shown to be impacted by sex differences, body mass index (BMI), and other factors.13) Accordingly, an appropriate renal function-estimating equation needs to be selected for each target patient.
In our previous study, we developed a modified CG equation geared toward improving the accuracy of estimation, particularly among older patients.14) In the present study, a modified CG equation was validated in a newly obtained cohort. In particular, we focused on patients whose estimation errors were large and found that body composition parameters, such as serum albumin levels, were statistically responsible for the discrepancy in CCr estimates. We incorporated the effect of body composition into the modified CG equation to improve the accuracy of the estimation.
The current study validated the modified CG equation, previously derived for an aged-oriented cohort, in a newly obtained dataset and analyzed the effect of altered body composition on this renal function-estimating equation.
PatientsThe height and weight of patients were measured by hospital staff on the day of CCr measurement. BMI15) and body surface area (BSA)16) were also calculated, and the measured CCr was adjusted using BSA. Considering the degree of obesity, body weight was modified to adjusted body weight (ABW).17,18)
Body composition was measured using the electrical impedance method with an InBody S20 or InBody 770 (BioSpace, Iowa, U.S.A.). Measured body composition parameters included muscle mass (kg), lean body mass (kg), body fat (kg), body fat (%), total body water (TBW; kg), intracellular water (kg), extracellular water (ECW; kg), edema value (ECW/TBW), protein (kg), bone mineral (kg), and body cell mass (kg).
CCr was measured using 24-h urine samples. Serum and urinary creatinine concentrations were measured using an enzymatic method commonly used in Japan.19)
Renal Function-Estimating EquationsIn the present study, we focused on the CG equation3) and modified CG equation developed for older patients in our previous study14) as follows:
*CG equation3)
 |
*Modified CG equation14)
 |
In addition, we focused on the MDRD equation20) to calculate eGFR from the input parameters of age, BUN, and albumin.
*MDRD equation20)
 |
Furthermore, we examined multiple regression equations with the CCr estimate as the dependent variable, calculated using the modified CG equation incorporating body composition parameters.
Factors Affecting CCr Estimates Using the Modified CG EquationA relative error (RE) was calculated using the following formula.
 |
The RE was calculated based on the measured CCr values. According to this value, patients were classified into two categories: |RE| ≥ 50% and |RE| < 10%.
Considering the two categories of |RE| ≥ 50% and |RE| < 10% for patients exceeding 65 years of age, factors affecting CCr estimates determined using the modified CG equation were predicted using the Mann–Whitney U test in terms of items of blood test results and body composition parameters. The cutoff value of this decisive factor for demarcating the two categories was determined using receiver operating characteristic (ROC) curve analysis. The area under the curve (AUC), sensitivity (%), and specificity (%) at optimal cutoff values were derived from the ROC curve.
Regression AnalysisUnivariate linear regression analysis was performed between the dependent variable of the measured CCr and one of the independent variables of the following body composition parameters: skeletal muscle mass (kg), lean fat mass (kg), body fat (kg), body fat (%), protein mass (kg), body cell mass (kg), total body water (kg), and ECW/TBW. Among these parameters, influential independent variables were determined for p < 0.05.
Multiple linear regression analyses were performed between the dependent variable of measured CCr and independent variables of CCr estimates using the modified CG equation (CCr-modified CG) and body composition parameters that fulfilled p < 0.05 in the univariable linear regression analysis.
A multivariable linear regression equation was derived considering the body composition parameters and CCr estimates using the modified CG equation, called the multivariate linear regression (MLR)-modified CG equation. A regression line was derived using the Deming regression method for a scatter plot of the measured CCr versus CCr estimates.
Statistical AnalysisStatistical analyses were performed using SPSS for Windows version 25.0 (IBM Corp., Armonk, NY, U.S.A.) and XLSTAT 2020.5.1 (Addinsoft, Paris, France). The Shapiro–Wilk test was used to assess the normality of the distribution of all parameters. Differences between the measured CCr and CCr estimates were analyzed using the following mean prediction error (ME%) and mean absolute prediction error (MAE%) indices21)
 |
Herein, the group of CCr estimates determined using the CG equation was called CCr-CG. Similarly, groups in which the CCr estimates were derived using the modified CG and MLR-modified CG equations were termed CCr-modified CG and CCr-MLR-modified CG, respectively. A paired t-test was applied to examine the ME (%) and MAE (%) of CCr-CG and CCr-modified CG or CCr-MLR-modified CG, respectively. In addition, two groups of CCr-modified CG and CCr-MLR-modified CG were evaluated using Dunnett’s test, with reference to CCr-CG as the control group. Furthermore, the percentage number of patients exceeding ≥65 years of age was derived for the two RE conditions (|RE| < 50% and |RE| < 10%). The chi-square test, in which the p-values were corrected using the Bonferroni method, was applied to test the CCr-modified CG and CCr-MLR-modified CG, using CCr-CG as a control group. The Spearman’s rank test was used to test the correlation coefficients. Statistical significance was defined as a two-sided p-value of <0.05.
The performance of the equations for estimating renal function was evaluated using the present dataset. The results were analyzed to identify factors affecting discrepancies between observed and estimated values.
Validation of Renal Function-Estimating EquationsPatientsIn total, 2601 patients (1528 males and 1073 females) were included in the dataset, and their clinical characteristics are summarized in Table 1. The dataset was divided into four groups according to age: Generation I, 258 patients (147 males and 111 females); Generation II, 1205 patients (697 males and 508 females); Generation III, 700 patients (416 males and 284 females); and Generation IV, 438 patients (268 males and 170 females).
| All (n = 2601) | Males (n = 1528) | Females (n = 1073) | Range | |
|---|---|---|---|---|
| Age (years) | 63 [19.00] | 63 [19.00] | 62 [19.00] | 20–93 |
| Height (m) | 1.62 [0.14] | 1.67 [0.09] | 1.54 [0.09] | 0.94–1.97 |
| Body weight (kg) | 60.1 [17.47] | 64.5 [16.17] | 52.8 [15.10] | 26.8–149.2 |
| ABW (kg) | 57.75 [17.23] | 63.85 [14.09] | 49.50 [10.80] | 21.69–112.40 |
| BSA (m2) | 1.63 [0.28] | 1.72 [0.22] | 1.49 [0.20] | 1.06–2.58 |
| BMI (kg/m2) | 23.03 [5.53] | 23.29 [5.15] | 22.66 [6.07] | 11.4–63.9 |
| Measured CCr (mL/min/1.73 m2) | 83.2 [51.40] | 82.2 [51.65] | 84.8 [50.85] | 1.3–334.5 |
| SCr (mg/dL) | 0.79 [0.40] | 0.87 [0.39] | 0.64 [0.29] | 0.20–6.29 |
| Urinary creatinine conc. (mg/mL) | 59.0 [52.0] (n = 435) | 66.5 [52.0] (n = 240) | 47.0 [40.0] (n = 195) | 9–309 |
| BUN (mg/dL) | 15.0 [7.00] (n = 2597) | 15.0 [8.00] (n = 1525) | 14.0 [7.00] (n = 1072) | 2–141 |
| Serum albumin (g/dL) | 3.70 [0.90] (n = 2409) | 3.80 [0.80] (n = 1433) | 3.70 [0.90] (n = 976) | 1.2–5.3 |
| Cystatin C (mg/L) | 1.23 [0.61] (n = 84) | 1.18 [0.53] (n = 50) | 1.29 [0.69] (n = 34) | 0.57–4.30 |
| CRP (mg/dL) | 0.23 [1.23] (n = 2155) | 0.31 [1.48] (n = 1316) | 0.14 [0.82] (n = 839) | 0.01–29.26 |
Data are presented as the mean and standard deviation (mean ± S.D.) for normally distributed variables and as the median and interquartile range for non-normally distributed variables. ABW, adjusted body weight; BSA, body surface area; BMI, body mass index; CCr, creatinine clearance; SCr, serum creatinine; BUN, blood urea nitrogen; CRP, C-reactive protein.
The ME (%) and MAE (%) values between measured and estimated CCr were evaluated using conventional and modified CG equations, respectively. A paired t-test was applied to differences in MAE (%) values between the conventional CG and modified CG equations to evaluate p-values (Table 2). As seen in Tables 1–2 for Generations III and IV, the MAE (%) values calculated using the modified CG equation were significantly lower than those calculated using the conventional CG equation. Therefore, the accuracy of the modified CG equation was higher than that of the conventional CG equation for patients aged ≥65 years in the present dataset.
| CG | Modified CG | p-Value | ||||
|---|---|---|---|---|---|---|
| ME (%) | MAE (%) | ME (%) | MAE (%) | |||
| All (n = 2601) | −10.53 | 30.57 | 11.18 | 29.81 | 0.376 | |
| Generation | II–IV (n = 2343) | −12.56 | 30.77 | 10.24 | 29.17 | 0.086 |
| I (n = 258) | 7.83 | 28.72 | 26.80 | 35.62 | <0.01 | |
| II (n = 1205) | −5.12 | 31.37 | 16.77 | 33.40 | 0.229 | |
| III (n = 700) | −20.25 | 29.90 | 2.56 | 24.45 | <0.01 | |
| IV (n = 438) | −20.72 | 30.51 | 4.54 | 25.07 | <0.01 | |
CG, Cockcroft–Gault; ME%, mean prediction error; MAE%, mean absolute prediction error mean. The paired t-test was applied to the differences in MAE (%) values between the conventional and modified CG equations to calculate p-values.
To examine the effects of parameters on the predictability of the modified CG equation, each generation group of patients was divided into two categories: one with |RE| ≥ 50% and the other with |RE| < 10%. The |RE| ≥ 50% category was considered the group of patients in which an error for CCr prediction using the modified CG equation was relatively large, while the prediction was accurate for the |RE| < 10% category. The Mann–Whitney U test was applied to these two categories considering various blood test results, as well as items such as height, weight, BSA, and BMI. We noted a significant difference in serum albumin level (g/dL) between |RE| ≥ 50% and |RE| < 10% categories. Table 3 summarizes the median values of serum albumin for the generation groups and the results of the Mann–Whitney U test. The serum albumin level for |RE| ≥ 50% was significantly lower than that for |RE| < 10% over the entire generation group, although the p-value of Generation III (p = 0.095) was relatively higher than that of other groups.
| |RE| ≥ 50% | |RE| < 10% | p-Value | ||||
|---|---|---|---|---|---|---|
| Albumin(g/dL) | Generation II–IV | 3.30 [1.28] | (n = 208) | 3.70 [0.80] | (n = 623) | <0.01 |
| Generation II | 3.40 [1.20] | (n = 119) | 3.90 [0.80] | (n = 312) | <0.01 | |
| Generation III | 3.40 [1.20] | (n = 52) | 3.60 [0.80] | (n = 187) | 0.095 | |
| Generation IV | 2.70 [1.15] | (n = 37) | 3.50 [0.78] | (n = 124) | <0.01 | |
RE, relative error. Data are presented as medians and interquartile ranges. Each generation group was tested using the Mann–Whitney U test.
Using the MDRD equation, comprising serum albumin as an input variable, we verified whether the estimation accuracy of this equation was influenced by the range of the two categories (|RE| ≥ 50% and |RE| < 10%). |RE| values were evaluated using the modified CG equation, as described in the previous section. Subsequently, we compared ME (%) and MAE (%) values of the MDRD equation for patients classified as |RE| ≥ 50% and |RE| < 10% using the Mann–Whitney U test. As the MDRD equation estimates GFR, ME values (%), as shown in Table 4, were calculated by correcting the measured CCr values by a factor of 0.7. Herein, CCr was assumed to be 30% higher than GFR owing to creatinine secretion.14) As seen in Table 4, for each generation group evaluated using the Mann–Whitney U test, ME (%) and MAE (%) values for patients classified under the |RE| ≥ 50% condition were significantly higher than those of patients classified as |RE| < 10%. Thus, the predictive ability of the MDRD equation was reduced for patients with decreased serum albumin levels, similar to the modified CG equation.
| |RE| ≥ 50% | |RE| < 10% | p-Value | |||||
|---|---|---|---|---|---|---|---|
| ME (%) | MAE (%) | n | ME (%) | MAE (%) | n | ||
| Generation II–IV | 102.80 | 103.93 | 208 | 27.22 | 27.30 | 623 | <0.01 |
| Generation II | 113.73 | 115.42 | 119 | 23.96 | 24.03 | 312 | <0.01 |
| Generation III | 87.70 | 88.32 | 52 | 28.20 | 28.34 | 187 | <0.01 |
| Generation IV | 88.89 | 88.89 | 37 | 33.88 | 33.93 | 124 | <0.01 |
CG, Cockcroft–Gault; ME%, mean prediction error; MAE%, mean absolute prediction error; MDRD, Modification of Diet in Renal Disease; RE, relative error. Each generation group was tested using the Mann–Whitney U test.
Considering serum albumin, an optimum cutoff value was selected by producing a ROC curve and comparing the |RE| ≥ 50% and |RE| < 10% categories for patients over the age ≥ 65 years (Fig. 1). In addition, parameters of sensitivity (%), specificity (%), AUC, p-value, and 95% confidence interval (CI) were derived at the cutoff value (Table 5). Herein, the sensitivity (or specificity) was defined as the proportion of CCr estimates belonging to |RE| ≥ 50% (or |RE| < 10%) and low (or high) albumin levels. Accordingly, the absolute relative error |RE| of the modified CG equation exceeded 50% if the albumin level was below the cutoff value of 3.1 g/dL.
ROC, receiver operating characteristic.
| Cutoff value of serum albumin (g/dL) | Sensitivity (%) | Specificity (%) | AUC | p-Value | 95% CI |
|---|---|---|---|---|---|
| 3.10 | 77.4 | 50.6 | 0.655 | <0.05 | 0.590–0.720 |
AUC, area under the curve; CI, confidence interval; ROC, receiver operating characteristic.
In the previous section, we revealed that the predictability of the CCr-estimating equation was reduced owing to a decline in serum albumin levels. Given that serum albumin levels are related to body composition, an MLR equation was developed by incorporating body composition parameters and estimates using the modified CG equation.
PatientsWe selected 100 patients (54 males and 46 females) whose interval between CCr and body composition measurements was less than 60 d. This dataset, selected from the original dataset (Table 1), included body composition parameters and was used to develop a multivariable linear regression-modified CG (MLR-modified CG) equation incorporating body composition parameters (Table 6).
| All (n = 100) | Males (n = 54) | Females (n = 46) | Range | |
|---|---|---|---|---|
| Age (years) | 59.50 [17] | 62.00 [17] | 55.91 ± 12.21 | 25–82 |
| Body weight (kg) | 59.00 [25.13] | 61.75 [22.20] | 55.15 [33.00] | 26.8–125.2 |
| BSA (m2) | 1.62 [0.36] | 1.69 [0.26] | 1.52 [0.38] | 1.12–2.25 |
| BMI (kg/m2) | 23.20 [7.81] | 22.62 [7.00] | 23.59 [16.15] | 11.4–56.09 |
| ABW (kg) | 57.03 ± 12.58 | 61.54 ± 11.05 | 51.74 ± 12.31 | 26.8–91.8 |
| Measured CCr (mL/min/1.73 m2) | 73.50 [51.0] | 77.89 ± 35.20 | 81.22 ± 40.32 | 3.2–197.5 |
| SCr (mg/dL) | 0.78 [0.57] | 0.91 [0.47] | 0.60 [0.34] | 0.24–4.64 |
| Serum albumin (g/dL) | 3.34 ± 0.75 (n = 89) | 3.30 ± 0.78 (n = 51) | 3.39 ± 0.71 (n = 38) | 1.2–4.7 |
| BUN (mg/dL) | 15.00 [10] (n = 98) | 18.00 [11] (n = 53) | 13.00 [10] (n = 45) | 4–71 |
| Cystatin C (mg/dL) | 1.55 ± 0.63 (n = 5) | 1.68 ± 0.85 (n = 3) | 1.35 ± 0.085 (n = 2) | 0.72–2.35 |
| CRP (mg/dL) | 0.50 [1.67] (n = 83) | 0.50 [2.17] (n = 50) | 0.50 [1.11] (n = 33) | 0.01–19.33 |
| Skeletal muscle mass (kg) | 23.25 [9.88] | 26.61 ± 6.01 | 18.95 [8.17] | 12.1–43.0 |
| Lean fat mass (kg) | 44.58 ± 10.79 | 49.24 ± 9.98 | 36.05 [12.5] | 24.8–74.9 |
| Body fat (kg) | 15.00 [15.28] | 13.30 [10.15] | 20.10 [24.95] | 0.90–71.0 |
| Body fat (%) | 25.55 [19.45] | 22.52 ± 8.54 | 35.50 [25.98] | 3.0–59.6 |
| ECW/TBW | 0.40 [0.025] | 0.40 ± 0.014 | 0.41 ± 0.019 | 0.37–0.45 |
| Body cell mass (kg) | 27.75 [10.78] | 31.42 ± 6.60 | 23.00 [8.98] | 15.5–49.4 |
| Days difference (d) | 6.93 [20] | 5.99 [15] | 10.80 [24] | 0–54 |
Data are presented as the mean and standard deviation (mean ± S.D.) for normally distributed variables and as the median and interquartile range for non-normally distributed variables. BSA, body surface area; BMI, body mass index; ABW, adjusted body weight; SCr, serum creatinine; BUN, blood urea nitrogen; CRP, C-reactive protein; ECW/TBW, extracellular water/total body water ratio.
Using the measured CCr as a dependent variable, univariate linear regression analysis was performed on male and female cohorts in the selected dataset to identify independent variables that meet p < 0.05. Accordingly, ECW/TBW and body fat (%) were extracted as independent variables.
MLR Estimating EquationUsing extracted variables, an MLR estimation equation was developed based on the linear relationship between the measured CCr as a dependent variable and body fat (%), ECW/TBW, and an estimate (CCr-modified CG), determined using the modified CG equation, as independent variables. The coefficients of these independent variables were determined for male and female cohorts with p < 0.05 (Table 7). The MLR equations obtained for male and female patients were as follows:
| MLR-modified CG | Male patients (n = 54) | Female patients (n = 46) | ||||||
|---|---|---|---|---|---|---|---|---|
| Constant | CCrModified CG | Fat (%) | ECW/TBW | Constant | CCrModified CG | Fat (%) | ECW/TBW | |
| Non-standardization coefficient | 225.729 | 0.961 | −0.673 | −525.156 | 474.193 | 0.668 | −1.052 | −1025.331 |
| t-Value | 2.481 | 8.809 | −1.775 | −2.393 | 3.733 | 4.591 | −2.49 | −3.518 |
| p-Value | 0.017 | <0.01 | 0.082 | 0.021 | 0.001 | <0.01 | 0.017 | 0.001 |
CCr, creatine clearance; MLR, multivariate linear regression; CCr-modified CG, CCr value calculated using the modified CG equation; ECW/TBW, extracellular water/total body water.
·MLR-modified CG:
 |
The ME (%) and MAE (%) values were calculated for measured CCr and CCr estimates using the conventional CG, modified CG, and MLR-modified CG equations. Dunnett’s test was applied to differences in the MAE (%) values between the control conventional CG group and the respective modified CG and MLR-modified CG groups to derive p-values (Table 8). Although no significant differences were observed (p > 0.05), the MAE (%) value of the MLR-modified CG equation was lower than that of the CG equation.
| All | Males | Females | ||||
|---|---|---|---|---|---|---|
| ME (%) | MAE (%) | p-Value | ME (%) | MAE (%) | p-Value | |
| CG (control) | −4.62 | 31.04 | — | 12.55 | 49.42 | — |
| Modified CG | 12.58 | 29.26 | n.s. | 26.51 | 41.11 | n.s. |
| MLR-modified CG | 8.15 | 24.56 | n.s. | 19.61 | 38.92 | n.s. |
CCr, creatine clearance; CG, Cockcroft–Gault; ME%, mean prediction error; MAE%, mean absolute prediction error mean; MLR, multivariate linear regression; n.s., not significant (p > 0.05). The p-values were calculated using Dunnett’s test for differences in MAE (%) values between the CG and relevant estimating equation group.
Table 9 summarizes the percentage number of patients over 65 years in the |RE| < 30% category, in which RE was calculated using the CG, modified CG, and MLR-modified CG equations, respectively, relative to measured CCr values. Considering male patients ≥65 years of age, the percentage number of patients under |RE| < 30% increased significantly in the modified CG and MLR-modified CG groups compared with that in the control CG group. Female patients showed a similar trend, except for relatively high p-values. These results indicate the improved accuracy of the modified renal function equations.
| Over the age of 65 | Males | Females | ||
|---|---|---|---|---|
| |RE| < 30% | p-Value | |RE| < 30% | p-Value | |
| CG (control) | 31.58% | — | 36.26% | — |
| Modified CG | 78.95% | 0.012 | 72.73% | n.s. |
| MLR-modified CG | 89.47% | p < 0.01 | 63.64% | n.s. |
Three cases are shown, with RE calculated using the CG, modified CG, and MLR-modified CG equations considering measured CCr values. CCr, creatine clearance; CG, Cockcroft–Gault; ME%, mean prediction error; MAE%, mean absolute prediction error; MLR, multivariate linear regression; n.s., not significant (p > 0.05). The p-values were calculated using Dunnett’s test for differences in the percentage of relative error between the control CG and relevant estimating equation groups.
Figures 2 (males) and 3 (females) present regression lines representing the relationship between the measured CCr and CCr estimates using the respective estimating equations. A linear equation (y = x) is also presented as a reference for estimating accuracy.
(a) CG. (b) Modified CG. (c) MLR-modified CG. CCr, creatine clearance; CG, Cockcroft–Gault; MLR, multivariate linear regression.
(a) CG. (b) Modified CG. (c) MLR-modified CG. CCr, creatine clearance; CG, Cockcroft–Gault; MLR, multivariate linear regression.
Spearman’s rank correlation coefficients between the measured CCr and CCr estimates, calculated using the relevant estimating equation for male patients, were 0.647 (p < 0.01) for the CG, 0.703 (p < 0.01) for the modified CG, and 0.751 (p < 0.01) for the MLR-modified CG equations. Similarly, Spearman’s rank correlation coefficients for female patients were 0.567 (p < 0.01) for the CG, 0.596 (p < 0.01) for the modified CG, and 0.711 (p < 0.01) for the MLR-modified CG equations.
For both male and female patients, the estimation accuracy was better with modified CG and MLR-modified CG equations than that with the conventional CG equation. Considering the results of the MLR-modified CG equation, the effect of incorporating body composition was not notable for females when compared with that for males.
The relationship between serum albumin levels and ECW/TBW for male and female patients is shown in Figs. 4 (a) and (b), respectively. Considering both male and female patients, the ECW/TBW value decreased with increasing serum albumin levels.
(a) Males. (b) Females. ECW, extracellular water; TBW, total body water.
Based on the observed results, we propose a flow diagram for appropriately selecting an equation for renal function estimation (Fig. 5).
In the present study, using a newly obtained dataset, we evaluated a modified CG equation that was previously developed for aged-oriented cohorts. The results of the MAE (%) for the modified CG equation were significantly lower than those for the conventional CG equation, particularly in older generations III (age 65–74) and IV (age ≥75 years) (Table 2). Furthermore, in the cohort aged ≥65 years, we identified a subset of patients whose estimates determined using the modified CG equation were inadequate (|RE| ≥ 50%). To elucidate a potential cause for this discrepancy, we compared |RE| ≥ 50% and |RE| < 10% in terms of various blood test items and found a significant difference in serum albumin level (g/dL) between the two cohorts. Thus, we applied univariate linear regression analysis to these cohorts of measured CCr and identified influential parameters related to body composition. Incorporating these parameters, we established a multivariable linear regression estimation equation based on the modified CG equation.
Based on the observed results, a decrease in the serum albumin level could underlie the deviation from the estimates obtained using the modified CG equation. Moreover, a cutoff value for serum albumin, which is likely to affect renal function estimates determined with the modified CG equation, was found to be 3.1 g/dL (Fig. 1, Table 5). Decreased serum albumin levels have been observed in patients with liver disorders, renal failure, nephrotic syndrome, malnutrition, and pregnant females.22) Renal estimating equations may result in inappropriate dose predictions when designing a dosage regimen for patients with decreased serum albumin levels. In addition, caution should be exercised when considering increased unbound drug concentrations in these patients. Sarcopenia, the loss of muscle mass due to aging, disease, immobility, and malnutrition, is a known factor underlying a decline in albumin levels. Accordingly, sarcopenia-mediated changes in body composition may impact renal function estimates.7) Conversely, renal function should be carefully predicted, especially for obese patients using the MDRD equation, given that this equation includes SCr, which is affected by body composition, as a variable.9) Moreover, given that the accuracy of equations estimating renal function varies depending on BMI, the effect of body composition must be considered when estimating renal function.
According to a study examining the accuracy of renal function using BMI in patients with chronic kidney disease, obesity was found to impact the prediction of CCr calculated using the conventional CG equation.9)
Based on these results, the impact of altered body composition on renal function was analyzed in detail. First, using univariable linear regression analysis, we identified body fat (%) and ECW/TBW as factors affecting renal function estimates with the modified CG equation. In addition, according to the developed multivariable linear regression estimation equation incorporating these parameters, the coefficients for these variables were negative, indicating that increased body fat (%) and edema decreased the renal function estimates (Table 7).
Tai et al. have reported that an elevated ECW/TBW ratio is associated with reduced renal function and serum albumin levels.23) In the present study, a decrease in serum albumin levels was observed in patients with a high ECW/TBW value (Fig. 4). Therefore, it is likely that increases in body fat (%) and ECW/TBW are decisive factors in the degradation of renal function. The reference value of ECW/TBW, as an index of edema, ranges between 0.36 and 0.39, and edema is more likely when the value exceeds 0.40. Herein, because the coefficients for body fat (%) and ECW/TBW in the multivariable analysis were negative, the currently available equations for estimating renal function are inappropriate for patients with severe obesity and edema.
Although there was no significant difference in the MAE (%) values between the conventional CG equation and the MLR-modified CG equation, which was improved by incorporating the effects of body fat (%) and ECW/TBW, we observed that MAE (%) values declined on applying the modified CG and MLR-modified CG equations (Table 8). For male patients over the age of 65 years, a significant increase in the percentage of patients with a relative error |RE| < 30% was observed using the MLR-modified CG equation when compared with the conventional CG equation (Table 9). However, considering female patients in the present dataset, no significant difference in the MAE (%) was observed between the MLR-modified CG and conventional CG equations, although the MAE (%) value for the MLR-modified CG equation was lower than that for the conventional CG equation. One possible reason for the significant deviation from the ideal line in this female cohort could be the presence of a group of patients who were significantly younger and heavier than the other female patients.
There is no consensus on whether the progression of sarcopenia is impacted by sex; however, a higher incidence of sarcopenia has been associated with the loss of muscle mass in females.13) Postmenopausal females reportedly experience substantial loss of lean fat mass accompanied by fat accumulation.24,25) Menopause-induced hormonal changes in females could explain sex-based differences in renal function estimates with alterations in body composition. Further studies are required to confirm this hypothesis.
Muscle mass has been shown to impact SCr, a muscle metabolite and a variable in the CG equation. Therefore, muscle mass and lean fat mass can be used to correct renal function estimates.26) Herein, muscle mass had no statistically significant effect on renal function. However, as SCr is an important biomarker of renal function, further studies are needed to determine the effects of changes in body composition on SCr.
Cystatin C is one of the indicators of renal function in older people. While serum creatinine levels are affected by muscle mass, cystatin C levels are not affected by it. We could not perform this analysis because of insufficient data for statistical analysis.
The present study has some limitations. First, to obtain a large number of patients for the statistical analysis, the time interval between the measurement of CCr and body composition was extended to 60 d. For this reason, data from acute cases that affect body composition and serum creatinine levels, such as acute kidney injury and hemodynamic abnormality, could not be completely excluded. Second, the present cohort comprised relatively young subjects, particularly the female cohort. Given that the effect of body composition on renal function is evident in older individuals, a detailed study should be undertaken for older patients. Third, the cohort in our present study comprised hospitalized patients who were potentially at a higher risk of frailty. This fact should be taken into account when applying the present results to outpatients or healthy individuals.
We would like to thank the reviewers for valuable comments.
The authors declare no conflict of interest.
