An efficient approach for glomerular filtration rate assessment in older adults

Abstract

Aims

Assessment of glomerular filtration rate (GFR) is crucial because the GFR value defines the stage of chronic kidney disease and determines the adjustment of drug dosage. The aim was to investigate a new method for the accurate determination of GFR in older adults based on the combination of an exogenous filtration marker, iohexol, and an endogenous marker, serum creatinine or cystatin C.

Methods

We combined variables for the estimation of GFR with a reduced set of measurements of the marker iohexol. In a population-based sample of 570 subjects (≥70 years old) from the Berlin Initiative Study (BIS), we investigated the following: (i) the BIS1 and BIS2 equations based on age, gender and serum creatinine with or without serum cystatin C; (ii) equations based on one or two iohexol measurements; and (iii) equations based on the combination of variables from BIS1 or BIS2 with iohexol measurements. The reference standard was based on eight iohexol measurements. The cut-off value of 60 ml min⁻¹ (1.73 m)⁻² was chosen to assess accuracy. Equations were constructed using a learning sample (n = 285) and an independent validation sample (n = 285).

Results

Misclassification rates were 17.2% (BIS1), 11.6% (BIS2), 14.7% [iohexol measurement at 240 min (iohexol₂₄₀)], 7.0% (iohexol₂₄₀ combined with variables included in BIS1) and 6.7% (iohexol₂₄₀ combined with variables included in BIS2). Misclassification rates did not decrease significantly after inclusion of two or three iohexol measurements.

Conclusions

Combined strategies for the determination of GFR lead to a relevant increase of diagnostic validity.

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT

• Equations for the estimation of glomerular filtration rate (GFR) based on endogenous markers have limitations.

• Current protocols for measuring GFR based on exogenous markers require several measurement points and are rather complex and time consuming.

WHAT THIS STUDY ADDS

• The approach of combining exogenous (iohexol) with endogenous filtration markers (creatinine or cystatin C) is new.

• Only one iohexol measurement combined with endogenous markers is necessary to obtain a valid GFR.

• The combined approach may reduce effort considerably (time and personel; thus increasing feasibility) compared with the original protocols for measuring GFR and it may improve the accuracy of GFR estimation compared with the use of endogenous markers only.

Introduction

Assessment of glomerular filtration rate (GFR) is essential for diagnosis and prognosis of chronic kidney disease (CKD) but also for improved decision making to determine the correct dosage of medication or diagnostic imaging (using potentially nephrotoxic contrast agents). Especially in elderly persons, in whom GFR may be more susceptible to medical treatment, careful dosage of medication is crucial to avoid further deterioration of kidney function. Unfortunately, most of the estimated glomerular filtration rate (eGFR) equations are not validated in older adults. Worldwide, the most frequently used equation is the Modification of Diet in Renal Diseases (MDRD) equation, which leads to an overestimation of GFR in elderly individuals [1] and makes it a problematic tool for correct drug dosage calculation. The GFR can also be measured (mGFR) using exogenous filtration markers, such as iothalamate or iohexol, by comparing an intravenous dose with subsequently measured concentrations in the serum or urine or both. In many cases, however, the GFR is estimated (eGFR) from concentrations of endogenous metabolites, such as serum creatinine or cystatin C.

This study is a secondary analysis of data from the Berlin Initiative Study (BIS). In a cross-sectional analysis of the BIS [2], two new equations for eGFR were recently developed using prospectively measured data from an elderly Caucasian population (age 70–96 years) [3]. Both equations included age, sex and serum creatinine, with one equation additionally including serum cystatin C. The reference standard was obtained from eight iohexol concentration measurements in a time window of 30–300 min after injection.

In contrast to clinical or experimental studies, eight measurements within 5 h are not feasible for clinical routine. The question remains whether it is possible to use a reduced number of measurements to obtain an mGFR more accurate than the eGFR estimated by equations such as the currently published BIS1, BIS2 or the Chronic Kidney Disease-Epidemiology Collaboration (CKD-EPI) equation [4]. Additionally, the time interval after injection to yield valid measurements is unknown.

Given that information on age, sex, bodyweight and laboratory parameters, including serum creatinine and, in some cases, serum cystatin C, is routinely available, this information could be added to a reduced number of iohexol measurements to improve assessment of GFR. Thus, our approaches combine estimation and (reduced) measurement of GFR, which is termed ‘emGFR’. We do not investigate approaches including cystatin C but not creatinine, because this situation is rarely met in daily practice. The ‘benchmark’ for determining the emGFR is the accuracy of eGFR as established by existing equations, such as BIS1, BIS2 or CKD-EPI.

In our study, we investigate strategies for the determination of emGFR, focusing on the clinical cut-off value of 60 ml min⁻¹ (1.73 m)⁻². The strategies differ in the number and time points of iohexol measurements. We restrict our analysis to a maximum of three iohexol measurements.

Methods

The study population and methods of obtaining and processing iohexol measurements have been described in detail previously [2,3]. Here we give a short summary of the methods applied.

Study population

The study population consisted of 570 Caucasians aged ≥70 years with valid iohexol measurements. This was a planned, population-based subsample of 2073 older adults recruited in the BIS, a cohort study that investigates the epidemiology of chronic kidney disease in old age. All participants had given informed consent for 5 h of iohexol clearance measurement. Apart from iohexol results, we use sociodemographic and anthropometric data, as well as results of blood parameters (creatinine and cystatin C). The ethics committee of the Charité – Universitätsmedizin Berlin approved the study.

Iohexol measurements

Participants with a thyroid-stimulating hormone level <0.3 mIU l⁻¹ or a known iodine allergy were excluded. Iohexol solution (5 ml, containing 3235 mg of iohexol; Accupaque; GE Healthcare Buchler, Braunschweig, Germany) was administered intravenously into an antecubital, forearm or hand vein and flushed with 10 ml of saline. Blood samples were obtained from the contralateral arm every 30 min for 240 min (4 h) and after 300 min (5 h) for a last time. Samples were centrifuged for 10 min at 4000g within 2 h of collection and transported on dry ice to be stored at −80°C until further analysis at Charité University Hospital.

Samples were assayed by high-performance liquid chromatography (HPLC). The HPLC analysis of the supernatant was carried out on a Hitachi HPLC system with a Diode array detector (Hitachi) and a Chromolith performance HPLC column (RP-18e; 100 mm × 4.6 mm; Merck) and a Chromolith guard-column (RP-18e; 5 mm × 4.6 mm; Merck) [5,6]. The GFR was calculated with the clearance computed from the amount of the marker administered and the area under the curve (AUC) of plasma concentration vs. time. This was estimated using a two-compartment model with early (three time points until 90 min; fast component) and late blood sampling (five time points from 120 min onward; slow component) [6]. The measured GFR and, consequently, all GFR estimating equations were corrected for body surface area by the factor individual body surface area/1.73.

Serum measurements

All creatinine samples were analysed in the same laboratory (Synlab MVZ Heidelberg GmbH) using the isotope-dilution mass spectrometry (IDMS) traceable enzymatic method from Roche (Crea plus; Roche Diagnostics, Mannheim, Germany) on a Roche Modular-analyzer P-Modul. Creatinine concentrations are given in milligrams per decilitre. To convert milligrams per decilitre to micromoles per litre, values are to be multiplied by 88.4.

Cystatin C samples were sent to the Charité laboratory, ‘Labor Berlin’, and measured using a particle-enhanced nephelometric assay (PENIA) on the BN-ProSpec nephelometer (Siemens Health Care Diagnostics, ex-Dade-Behring, Marburg, Germany). The manufacturer's reference interval for healthy subjects is 0.59–1.05 mg l⁻¹ (after restandardization of cystatin C according to ERM – DA 471/IFCC for BN Systems). All cystatin C samples were frozen at −80°C and analysed in one batch within 4 days. Cystatin C is known to be stable at −80°C [7]. More details on coefficients of variation and calibration were given previously [3].

Statistical methods

The aim of the statistical analysis was to determine how many iohexol measurements are necessary and which time points are optimal to obtain a valid emGFR.

We developed the proposed strategies in a learning sample and validated their performance in a validation sample. To avoid overfitting, equations for one measurement point were constructed without using this measurement point in the gold-standard mGFR. Thus, for the equations including iohexol measurements at 240 min, mGFR was determined using iohexol measurements at 30, 60, 90, 120, 150, 180 and 300 min but not at 240 min. For equations including measurements at 300 min, the mGFR was determined using iohexol measurements at 30, 60, 90, 120, 150, 180 and 300 min but not at 300 min. This approach, however, was not used for the description of the sample (Table 1) and the figures, but only for the evaluation of equations (Tables 2 and 3). For Table 1 and the figures, the mGFR based on eight iohexol measurement points (standard approach) was used. Equations were constructed using a double-logarithmic linear regression approach [6]. The single samples of iohexol were treated as covariates in the same way as the serum creatinine and cystatin C measurements. Variable selection was applied to a small number of variables set in advance, including iohexol measurements, age, gender, bodyweight, serum creatinine and cystatin C. In the validation sample, the model chosen on the basis of the learning sample was applied and an unadjusted correlation coefficient (r²) was reported. Without affecting these calculations, the coefficients for the equations presented were determined using all 570 subjects.

Table 1.

Description of study population by measured glomerular filtration rate

Characteristics	Total	mGFR ≥60 ml min−1 (1.73 m)−2*	mGFR of 30–59 ml min−1 (1.73 m)−2*	mGFR <30 ml min−1 (1.73 m)−2*
Participants [n (%)]	570 (100)	297 (52.1)	256 (44.9)	17 (3.0)
Age [years; mean (range)]†	78.5 (70–96)	76.6 (70–94)	80.4 (70–96)	84.3 (73–94)
Female [n (%)]	244 (42.8)	136 (45.8)	102 (39.8)	6 (35.3)
Diabetes mellitus [n (%)]‡	137 (24.0)	67 (22.6)	64 (25.0)	6 (35.3)
Height [m; mean (range)]	1.66 (1.43–1.92)	1.66 (1.43–1.92)	1.66 (1.46–1.89)	1.65 (1.52–1.78)
Weight [kg; mean (range)]	77.3 (47–136)	77.5 (47–136)	77.4 (48–131)	74.0 (53–103)
mGFR [ml min−1 (1.73 m)−2; mean (range)]	60.3 (15.5–116.7)	72.8 (60.0–116.7)	48.2 (30.3–59.9)	23.9 (15.5–29.6)
Serum creatinine level [mg dl−1; mean (range)]§	0.99 (0.46–4.77)	0.81 (0.46–1.19)	1.13 (0.52–2.47)	2.12 (1.12–4.77)
Serum cystatin C level [mg l−1; mean (range)]¶	1.15 (0.61–4.40)	0.93 (0.61–1.44)	1.31 (0.82–2.36)	2.37 (1.38–4.40)
Area under concentration–time curve for iohexol [(ml)/1.73 m2; mean (range)]**	55.7 (22.6–219.1)	42.6 (22.6–63.1)	65.5 (40.0–116.2)	136.0 (99.9–219.1)
GFR based on 240 min concentration only [ml min−1 (1.73 m)−2; mean (range)]††	64.6 (17.6–131.6)	76.7 (53.2–131.6)	53.0 (29.1–81.0)	28.1 (17.6–42.9)

^*mGFR is the measured glomerular filtration rate using eight iohexol measurements.

^†For calculation of means, age was calculated exactly in days, whereas ranges refer to complete years, i.e. truncation of exactly calculated values.

^‡Diabetes was defined as either haemoglobin A_1c >6.5% or prescription of antidiabetic medication.

^§To convert creatinine from milligrams per decilitre to micromoles per litre, multiply by 88.4.

^¶Standardized cystatin C values were converted by formula (−0.105 + 1.13 × cystatin C) before being used for Equation 3 in the manuscript. To convert cystatin C from milligrams per litre to nanomoles per litre, multiply by 74.9. More details can be found in Schaeffner [3].

^**The area under the concentration–time curve was calculated by the method of Schwartz et al. [6] using a double-exponential model with fast and slow components.

^††The glomerular filtration rate (GFR) was calculated according to the equation of Bird et al. [16] using iohexol concentration, bodyweight and gender.

Table 2.

Bias and correlation (r²) in the validation sample (n = 285) for selected emGFR equations in persons aged ≥70 years

Equation	Mean bias*	SD of differences	r2	Wrongly considered <60 ml min−1 (1.73 m)−2 [n (%), 95% CI]	Wrongly considered >60 ml min−1 (1.73 m)−2 [n (%), 95% CI]	Total misclassified [n (%), 95% CI]	P values†
Routine variables‡ [no iohexol (BIS1)]	0.11	9.20	0.75	27 (9.5)	22 (7.7)	49 (17.2)	NA
				(6.3–13.5)	(4.9–11.5)	(13.0–22.0)
Routine variables + cystatin C [no iohexol (BIS2)]	0.09	8.06	0.82	18 (6.3)	15 (5.3)	33 (11.6)	NA
				(3.8–9.8)	(3.0–8.5)	(8.1–15.9)
Iohexol after 180 min	-0.69	8.40	0.74	25 (8.8)	19 (6.7)	44 (15.4)	0.51
				(5.8–12.7)	(4.1–10.2)	(10.8–19.4)
+ routine variables	0.55	5.15	0.90	13 (4.6)	12 (4.2)	25 (8.8)	<0.001
				(2.5–7.7)	(2.2–7.2)	(5.8–12.7)
+ routine variables + cystatin C	0.42	4.97	0.92	14 (4.9)	8 (2.8)	22 (7.7)	0.02
				(2.7–8.1)	(1.2–5.5)	(4.9–11.5)
Iohexol after 240 min	−0.36	6.79	0.83	28 (9.8)	14 (4.9)	42 (14.7)	0.35
				(6.6–13.9)	(2.7–8.1)	(10.8–19.4)
+ routine variables	0.55	4.69	0.92	10 (3.5)	10 (3.5)	20 (7.0)	<0.001
				(1.7–6.4)	(1.7–6.4)	(4.3–10.6)
+ routine variables + cystatin C	0.46	4.59	0.93	9 (3.2)	10 (3.5)	19 (6.7)	0.004
				(1.5–5.9)	(1.7–6.4)	(4.1–10.2)
Iohexol after 300 min	−0.65	6.48	0.85	21 (7.4)	9 (3.2)	30 (10.5)	0.009
				(4.6–11.0)	(1.5–5.9)	(7.2–14.7)
+ routine variables	0.16	5.29	0.91	14 (4.9)	7 (2.5)	21 (7.4)	<0.001
				(2.7–8.1)	(1.0–5.0)	(4.6–11.0)
+ routine variables + cystatin C	0.09	5.19	0.91	16 (5.6)	8 (2.8)	24 (8.4)	0.08
				(3.2–9.0)	(1.2–5.5)	(5.5–12.3)

Abbreviation is as follows: CI, confidence interval; NA, not applicable.

^*Bias was defined as the difference between estimated/measured GFR (emGFR) and measured GFR (mGFR). Mean and SD refer to these differences.

^†P values refer to the sign test (two-sided) for the reduction of misclassification rates using the iohexol measurement only (shaded rows) or additional to iohexol measurement in comparison to the BIS1 equations. Models including only iohexol or routine variables plus iohexol are compared with the BIS1 equation, while models using routine variables plus cystatin C plus iohexol are compared with the BIS2 equation. Results for BIS1 and BIS2 have been published in [3]. For example, the misclassification using routine variables plus iohexol after 240 min (row 8, 42/285, 14.7%) was significantly smaller (P < 0.001) in comparison to using routine variables only (row 2, BIS1 equation, 49/285, 17.2%).

^‡Routine variables are age (in years), bodyweight (in kilograms) and serum creatinine level (in micromoles per litre). Gender was not included in the models because it modified the emGFR value by <1%.

Table 3.

Bias, correlation (r²) and misclassification rates in the validation sample (n = 285) for the emGFR equations in persons aged ≥70 years, with comparisons of one vs. two vs. three iohexol measurement points

Time points of iohexol measurement in the equation	Mean bias*	SD of differences	r2	Wrongly considered <60 ml min−1 (1.73 m)−2 [n (%), 95% CI]	Wrongly considered >60 ml min−1 (1.73 m)−2 [n (%), 95% CI]	Total misclassified [n (%), 95% CI]	P values†
180 min	−0.69	8.40	0.74	25 (8.8)	19 (6.7)	44 (15.4)	NA
				(5.8–12.7)	(4.1–10.2)	(10.8–19.4)
180 min + 120 min	−0.67	7.29	0.81	23 (8.1)	15 (5.3)	38 (13.3)	0.22
				(5.2–11.9)	(3.0–8.5)	(9.6–17.8)
180 min + 120 min + 60 min	−0.26	8.24	0.80	23 (8.1)	15 (5.3)	38 (13.3)	0.26
				(5.2–11.9)	(3.0–8.5)	(9.6–17.8)
240 min	−0.36	6.79	0.83	28 (9.8)	14 (4.9)	42 (14.7)	NA
				(6.6–13.9)	(2.7–8.1)	(10.8–19.4)
240 min + 180 min	−0.22	6.53	0.85	24 (8.4)	12 (4.2)	36 (12.6)	0.16
				(5.5–12.3)	(2.2–7.2)	(9.0–17.1)
240 min + 180 min + 60 min	−0.05	6.56	0.86	20 (7.0)	15 (5.3)	35 (12.3)	0.16
				(4.3–10.6)	(3.0–8.5)	(8.7–16.7)
300 min	−0.65	6.48	0.85	21 (7.4)	9 (3.2)	30 (10.5)	NA
				(4.6–11.0)	(1.5–5.9)	(7.2–14.7)
300 min + 240 min	−0.59	6.36	0.85	20 (7.0)	9 (3.2)	29 (10.2)	0.56
				(4.3–10.6)	(1.5–5.9)	(6.9–14.3)
300 min + 240 min + 60 min	−0.10	6.79	0.88	16 (5.6)	8 (2.8)	24 (8.4)	0.18
				(3.2–9.0)	(1.2–5.5)	(5.5–12.3)

Abbreviation is as follows: CI, confidence interval; NA, not applicable.

^*Bias was defined as the difference between estimated/measured GFR (emGFR) and measured GFR (mGFR). Mean and SD refer to these differences.

^†P values refer to the sign test (two-sided) for the comparison of misclassification rates of more than one measurement point vs. one measurement point in the shaded row above the selected row. For example, the total misclassification using 180, 120 and 60 min measurements (row 4, 38/285, 13.3%) was not significantly smaller (P = 0.26) than the total misclassification using the iohexol measurement at 180 min only (row 2, 44/285, 15.4%).

True classification rates for a cut-off value of 60 ml min⁻¹ (1.73 m)⁻² are the primary outcome. Bias was assessed as the mean difference between emGFR and mGFR, with positive values indicating an overestimation of mGFR, and accuracy as r², relative to mGFR. Construction and validation of the equations was done using SPSS for Windows 19.0 (IBM, Armonk, NY, USA).

Results

Study sample

Table 1 presents a short overview of the most important characteristics of the study population. More detailed information about the study sample can be found elsewhere [2,3]. Mean age was 78.5 years (range 70–96 years) with slightly more males (57.2%) than females. The mGFR was 60.3 ml min⁻¹ (1.73 m)⁻² (range 15.5–116.7 ml min⁻¹ (1.73 m)⁻²). About half of the subjects had an mGFR >60 ml min⁻¹ (1.73 m)⁻² and only 17 subjects (3%) had an mGFR <30 ml min⁻¹ (1.73 m)⁻². None of the 570 participants experienced an allergic reaction to iohexol or any other serious adverse event.

Equations for emGFR

Table 2 displays performance measures of the approaches using only one measurement point of iohexol alone and combined with different clinical routine parameters. Table 3 demonstrates the same for the use of two and three measurement points of iohexol but not including routine variables. The comparison of both tables shows that one late iohexol measurement is sufficient to obtain an accurate value of emGFR. An additional earlier iohexol measurement does not improve the accuracy as long as the later measurement is undertaken after 240 or 300 min. Table 2 shows that the optimal measurement point is 240 min. Thus, we propose the following three equations using the 240 min iohexol measurement (iohexol₂₄₀) value combined with different clinical routine parameters:

Equation 1

Equation 2

Equation 3

Units for emGFR are millilitres per minute per 1.73 m²; for iohexol, micrograms per millilitre; for age, years; for bodyweight, kilograms; for creatinine, milligrams per decilitre; and for cystatin C, milligrams per litre. Standardized cystatin C values were converted by formula (−0.105 + 1.13 × cystatin C) before being used for Equations 2 and 3 [8].

Gender is not included in the equations because correction factors were not significant and between 0.995 and 1.004, modifying the emGFR by <1%. Note that combined with one iohexol measurement, bodyweight becomes a significant predictor of emGFR. This is in contrast to the pure estimation of eGFR, as has been shown for the BIS1 and BIS2 equations [3]. Bias and accuracy for these three equations can be found in rows 7–9 of Table 2.

Figure 1 displays the accuracy of one iohexol measurement in terms of r². The bottom curve shows the accuracy of iohexol measurements alone at several time points, the middle curve combines iohexol measurements with clinical routine parameters (age, sex, bodyweight and serum creatinine), and the top curve includes information of an additional cystatin C measurement. The horizontal lines show the accuracy of eGFR based on creatinine measurements (dot-dashed line; BIS1 equation) and the combination of creatinine and cystatin C without iohexol measurements (continuous line; BIS2 equation). The figure clearly shows that if only one iohexol measurement is intended, the time point should be no earlier than 120 min after injection. If iohexol measurements 180 min after injection or later are used, the additional information provided by the creatinine measurement remains substantial, whereas the additional measurement of cystatin C does not seem to be necessary. This is confirmed by the results displayed in Tables 2 and 3, in which the different approaches are evaluated using the diagnostically relevant cut-off value of 60 ml min⁻¹ (1.73 m)⁻².

Figure 1.

The figure displays the accuracy of iohexol measurements in terms of r² (logarithmic scale) for prediction of measured glomerular filtration rate (mGFR) for different time points of measurement. The use of iohexol only is compared with the combination of iohexol measurements with established serum markers. The horizontal lines show the accuracy of estimated glomerular filtration rate (eGFR) based on creatinine (dot-dashed line, BIS1 equation) and the combination of creatinine and cystatin C without iohexol measurements (continuous line, BIS2 equation). () Iohexol + routine variables + cystatin C; () iohexol + routine variables, () iohexol only

In Figure 2 the percentage differences of emGFR and mGFR are displayed for Equations 1 and 2. The graph for Equation 3 looks very similar (not shown). It can clearly be seen that the equations show worse performance for very low values of mGFR if differences are calculated as percentages. In particular, when choosing a cut-off value of 30 mg ml⁻¹ (1.73 m)⁻², only 41.2% of n = 17 individuals below this cut-off value were correctly classified using Equation 1. Equations 2 and 3 both revealed a true classification rate of 58.8% for subjects below this cut-off value. In contrast, when choosing a cut-off value of 50 mg ml⁻¹ (1.73 m)⁻², this rate increased to 96.6% for both Equations 2 and 3. Figure 3 shows to some extent, but less strictly, that the unfavourable differences relate to high creatinine and high cystatin C values. It might be of interest that equations including cystatin C instead of creatinine measurements revealed comparable results in terms of accuracy and precision. In particular, the r² value of the models including iohexol (240 min), patient age and bodyweight were identical whether cystatin C only or creatinine values only were included (model r² in the validation sample = 0.93 and 0.92).

Figure 2.

The figure displays the differences between emGFR and mGFR (y-axis) vs. mGFR (x-axis) for Equation 1 [iohexol measurement at 240 min only (iohexol₂₄₀)] and Equation 2 (iohexol240 plus routine measurements including creatinine) in the total sample. The gain of accuracy using additional routine measurements can be seen clearly. The proportion of subjects with difference eGFR-mGFR at most 30% (P30) was 96.5% for Equation 1 and 97.2% for Equation 2 in the validation sample. All deviations of at least 30% refer to an overestimation of mGFR via emGFR for small values of mGFR

Figure 3.

The figure displays the differences between emGFR and mGFR (y-axis) vs. creatinine (x-axis of left panel) and cystatin C (x-axis of right panel) in the total sample. In accordance with Figure 2, deviations of ≥30% occur primarily for large values of creatinine and cystatin C

In summary, the results show that one late iohexol measurement combined with routine variables is a sufficient strategy for determining emGFR. In terms of bias, even the singular iohexol measurement in Equation 1 shows very good results (Table 2). However, accuracy in terms of correct classification rates according to the cut-off value of 60 ml min⁻¹ (1.73 m)⁻² is considerably improved by combining iohexol and clinical routine parameters.

Discussion

Clearance measurement by exogenous filtration markers is complex and costly. It is well known that further research is necessary to explore the validity of equations for determining GFR from endogenous markers in the elderly [9]. However, in several clinical situations, such as the evaluation of (older) kidney donors or the initiation of dialysis, but most often for dosing of chemotherapy and all other drugs that are renally cleared, an exact clearance measurement based on exogenous markers instead of GFR-estimating equations may be indicated. It is well known that the most frequently used MDRD equation is not an appropriate tool to calculate drug dosage in elderly patients [10]. In these clinical scenarios, it would be advantageous to have a procedure at hand that is more exact than GFR results obtained by estimating equations but of less complexity than full clearance measurements.

In a sample of 570 older adults of the BIS, we could demonstrate that the combination of clinical routine variables and one single iohexol measurement (240 min) reduced GFR misclassifications by 60% compared with routine variables without iohexol. We could show that the combined iohexol–routine variable approach may be useful as a confirmatory test for CKD in older adults, and the principal idea of the combination of eGFR and a reduced mGFR in one equation may also be useful in younger persons or the general population.

The usefulness of iohexol measurements could not be proved for time points earlier than 180 min after injection. This is in agreement with previous work that proposes to restrict sampling to late measurements, i.e. the ‘slow component’ [11,12]. A few studies have compared single-sample values based on one marker with a different reference method based on another marker [13–15]. We found only few studies directly comparing multi-sample and single-sample GFR using the identical filtration marker [16,17]. To use only a single sample instead of several (even after 4 h) is advantageous and cost saving. The injection of iohexol can be done in an outpatient setting and is known to be safe if individuals with hyperthyroidism or iodine allergy are excluded [18–21]. The idea of reducing the number of measurements for mGFR to one has been elaborated previously [16] based on measurements of ⁵¹Cr-EDTA and iohexol. Van Pottelbergh et al. recently published the design of a study protocol of one to three iohexol measurements to be used for subjects of all ages [22]. However, the idea of reducing measurement points has been mentioned implicitely as early as in the paper by Brochner-Mortensen [6,11]. Yet, the idea of adding routine variables to a single iohexol measurement in order to obtain more exact GFR results but avoid multiple measurements is, to the best of our knowledge, unique.

One possible explanation for the gain in accuracy by adding routine variables is that the limitation of the single iohexol measurements, especially at shorter sampling times, is due to an overestimate of the slope in the secondary clearance phase. The addition of creatinine may allow some correction for this.

There were two unexpected findings in the process of equation development. First, gender did not have a significantly predictive impact on eGFR compared with existing equations that generally adjust for gender. It can only be speculated that gender differences are mirrored in the measured iohexol concentration directly and thus do not have to be corrected for. A second noteworthy result was the fact that the prediction potential of the combination of iohexol and creatinine as well as routine variables could not be improved by adding cystatin C. Even though cystatin C is known to be a better marker than creatinine, especially in older adults, in whom changes in muscle mass are common [7,23,24], its addition did not lead to a significant improvement in predicting GFR when iohexol was present. However, in a larger sample an additional benefit of including this marker cannot be excluded. Due to the restricted availability and high cost of cystatin C measurement, we did not combine iohexol with routine variables and cystatin C excluding creatinine. However, we note that discrepant results of creatinine- and cystatin C-based eGFR (e.g. difference >40%) as well as eGFR measurements close to the cut-off value of 60 ml min⁻¹ (1.73 m)⁻² might favour additional iohexol measurements, leading either to a full mGFR using all eight measurement points [25] or to an emGFR similar to our approach. According to the approach of Björk et al., it might be useful to choose earlier or later iohexol measurement time points depending on the known value of eGFR based on creatinine or cystatin C or both [26].

Our algorithm differentiates between a GFR ranging around 60 ml min⁻¹ (1.73 m)⁻², which is the clinical decision-leading cut-off value, and values clearly below or above that cut-off value. However, similar approaches can also be developed for smaller cut-off values, which may be more relevant in older adults. In this sense, we interpret our study as a ‘proof of concept’, and different criteria may lead to slightly different equations in other settings. In particular, similar strategies might be developed for filtration markers other than iohexol, e.g. iothalamate. However, we believe that the main result, that one late iohexol measurement considerably improves the estimation of the GFR, holds in general.

One of the strengths of our study is the originality of combining the measurement of an exogenous filtration marker with endogenous laboratory and routine variables for GFR determination. Using only one measurement of iohexol demands far fewer personnel than several measurements. Further advantages are the primary data collection of a relatively large sample size and measurements of iohexol clearance that were done according to a detailed and consistent protocol and analysed in a very short period of several days. Besides, creatinine and cystatin C assays were carried out according to the highest standards.

Our work also has limitations. This was a secondary exploratory analysis, because the primary aim of the BIS project had been the construction of the BIS equations. Moreover, we only simulated strategies at the level of pre-existing data; we could not investigate the practical problems arising in clinical routine when implementing these strategies. In contrast to other studies mentioned above [10–12], we used the same filtration marker, iohexol, as the gold standard and for the reduced measurements. However, the BIS1 equation, which was developed using iohexol as the gold standard, has been externally validated in several independent studies [27,28] using a different gold standard, namely inulin, demonstrating excellent performance of the BIS1 equation. Additionally, using several statistical approaches (see Methods) we tried to avoid overfitting. Finally, the population of the BIS consisted of Caucasians with mild to moderately reduced kidney function. This might explain why our equations do not perform as well below a cut-off value of 30 ml min⁻¹ (1.73 m)⁻² (see Results). Possibly, a modification of these equations might be necessary if data from subjects with severely reduced kidney function are the focus. However, we cannot completely exclude the possibility that these differences are due to measurement errors of our gold standard, iohexol (mGFR with eight measurement points), especially if creatinine and cystatin C values are consistent.

Thus, an extension of our results to other ethnicities or patients with more severely reduced kidney function should be considered with caution.

In conclusion, we have shown that existing equations for eGFR may benefit from the addition of a single iohexol measurement. The combination of approaches, determining eGFR and mGFR with a reduced number of measurements, is a new and efficient approach for determining GFR precisely in clinical routine.

Competing Interests

All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any organization for the submitted work; no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years; no other relationships or activities that could appear to have influenced the submitted work.