• L. Abbigail Granger,

    Corresponding author
    • From the Department of Clinical Sciences, College of Veterinary Medicine, Kansas State University, Manhattan
    Search for more papers by this author
  • Laura J. Armbrust,

  • David C. Rankin,

  • Ronette Ghering,

  • Nora M. Bello,

  • Kate Alexander

  • Funding: Mentored Clinical, Applied, or Translational Research Program, College of Veterinary Medicine, Kansas State University

Address correspondence and reprint requests to Dr. Granger. E-mail:


Commonly used clinical indicators of renal disease are either insensitive to early dysfunction or have delayed results. Decreased glomerular filtration rate (GFR) indicates renal dysfunction before there is a loss of 50% of functional nephrons. Most tests evaluate global rather than individual kidney function. Dynamic computed tomography (CT) and Patlak plot analysis allows for individual GFR to be tested. Our objectives were to establish a procedure and provide reference values for determination of global GFR in 10 healthy cats using dynamic CT (CTGFR). This method of GFR determination was compared against serum iohexol clearance (SIC). A single CT slice centered on both kidneys and the aorta was acquired every fifth second during and after a bolus injection of iohexol (240 mgI/ml; 300 mgI/kg) for 115 s. Using data from this dynamic acquisition, Patlak plots were obtained, GFR was calculated, and results were compared to global GFR determined by iohexol clearance. The average global CTGFR estimate was 1.84 ml/min × kg (SD = 0.43; range = [1.22, 2.45]). The average global GFR measured using SIC was 2.45 ml/min × kg (SD = 0.58; range = [1.72, 3.69]). GFR measurements estimated by both dynamic CT and SIC were positively associated (estimated Spearman rank correlation coefficient = 0.72; P = 0.0234). The CTGFR method consistently underestimated GFR with a bias of −0.62 (SE = 0.1307) when compared to SIC (P = 0.0011). In healthy cats, CTGFR was capable of determining individual kidney function and appears clinically promising.


he prevalence of renal failure among cats is as high as 9.6% (96/1000 cats examined).[1] The prevalence increases to 20% in geriatric cats.[2, 3] Tests used to evaluate renal function include serum creatinine, blood urea nitrogen, and urine specific gravity. These tests are convenient, but insensitive until advanced functional loss of nephrons is present, especially in cats where concentrating ability may be maintained in spite of renal failure. [3-6] Therefore, additional methods are required to diagnose early renal dysfunction accurately.

Measurement of glomerular filtration rate (GFR) allows for early detection of renal dysfunction before overt failure is present.[7, 8] GFR is an ideal measure of renal function as it is directly proportional to the number and functional capacity of nephrons.[4, 9] Traditional estimation of GFR is accomplished by measuring clearance of a marker of known concentration by sampling urine or plasma. Nuclear scintigraphy is the only traditional method of GFR measurement that can determine individual kidney function. Markers used to estimate GFR must have a lack of binding to plasma protein, lack of reabsorption and secretion by renal tubules, limited metabolism, and free filtering through the glomeruli.[10, 11] Previously described markers include: insulin, creatinine, radiopharmaceuticals (most commonly, 99mTc-DTPA (diethylenetriamine-pentaacetic acid)), and iodinated contrast media.[4, 12-14] Inulin clearance is the gold standard for measurement of GFR, but the clearance of the iodinated contrast medium iohexol is an accurate alternative.[12, 13, 15-18] GFR measured using serum iohexol clearance (SIC) has good correlation with radionuclide-labeled markers and exogenous creatinine clearance.[4, 13, 15, 18]

A single injection of iohexol may be used to determine GFR accurately. Multiple methods have been described to measure serum concentration of iohexol, including x-ray fluorescence, alkaline hydrolysis and iodine measurement, capillary electrophoresis, and high performance liquid chromatography.[12, 17-19] Recently, inductively coupled plasma mass spectrometry (ICP/MS, Diagnostic Center for Population and Health—Michigan State University, Lansing, MI), has also been validated.[20] Using this method, serum samples are collected at predetermined times after injection of a known dose of iohexol. Concentrations of iohexol in the samples are used to estimate GFR.

Many tests used to estimate GFR are time consuming or have delayed results. Multiple urine or serum samples are often required and individual kidney function cannot be determined using most traditional methods of GFR estimation. Nuclear scintigraphy does allow for measurement of individual kidney function. None of these methods allow for determination of renal morphology. As a result, ultrasound or excretory urography is needed to assess renal parenchymal abnormalities.

Determination of GFR using computed tomography (CT) after injection of iohexol allows characterization of regional and global kidney function and provides excellent morphologic information in the same test. A benefit of CT is the ability to measure changes in concentration of contrast medium within virtually any region of interest (ROI) noninvasively with excellent spatial resolution.[10] A linear relationship exists between attenuation changes on CT and concentration of contrast medium in an ROI.[10] Likewise, the concentration of contrast medium in a given ROI and in a given organ is directly related to blood flow to that area. As renal filtration rate relies on renal blood flow, measuring changes in renal attenuation using CT after the injection of contrast medium allows quantification of renal function. Alterations in attenuation (Hounsfield units, HU) are directly proportional to GFR. Dynamic CT has been used to quantify GFR in humans, pigs, and dogs.[10, 21-24] The iohexol dose required for the GFR analysis using dynamic CT (CTGFR) is less than one-fourth of the dose required for standard CT contrast procedures and for serum analysis of iohexol as a tracer in humans, dogs, and pigs.[21, 22, 24] Many of these studies have found a close correlation between CTGFR and GFR measured by traditional means, but with underestimation of traditional reference values when measured by CTGFR. Validation of CTGFR in healthy cats has not been established. CTGFR has been used in cats with renal disease, but reference values were not available.[25] Our purpose was to establish a procedure and provide reference values for estimation of GFR using dynamic CT as compared to iohexol serum clearance in healthy cats using Patlak theory.

A Patlak plot is a graphical analysis technique based on pharmacokinetic principles of a two-compartment model developed originally for nuclear medicine.[26] The model requires that a tracer leaves one compartment (blood) and enters a second compartment irreversibly (filtrate).[10, 21, 24-31] Iohexol behaves in this manner in the kidney during the first 2 min after injection, before it exits via the ureters.[11] During this time, iohexol is accumulating within renal tubules and is leaving the vascular compartment. Given the direct relationship between iohexol concentration and CT attenuation measurements, the iohexol concentration in these compartments can be measured noninvasively by scanning an ROI sequentially and obtaining a time attenuation curve (TAC) of the kidneys and the aorta.

At a given time, t, the amount of iodine in a volume of kidney is represented by math formula. Iodine within the kidney is distributed between the renal blood volume, math formula, and the filtrate, math formula, such that math formula. The concentration of iodine in each compartment is obtained by dividing the amount of iodine in each of the previously described compartments by volume, V, such that math formula, math formula, and math formula. During the first 2 min after iohexol injection, the clearance of iodine follows a two-compartment model and the rate of change of iodine within the filtrate is proportional to the rate of change in blood. Factoring in the proportionality constant, α, which denotes rate of elimination of contrast medium, makes changes within each of the two renal compartments equivalent, such that math formula. By definition, the proportionality constant, α, represents whole blood clearance. Integrating this equation results in:

display math(1)

As mentioned previously, math formula is the amount of iodine in a volume of kidney and the concentration of iodine in the kidney is math formula. Therefore, math formula. Dividing both sides by math formula, the ratio of iodine concentration in the kidney to blood ratio of iodine concentration is obtained and results in math formula. Substituting this equation with equation (1) results in:

display math(2)

The relationship, math formula, is equivalent to blood volume, math formula is equivalent to fractional vascular volume (fvv). Finally,

display math(3)

Graphically, equation (3) represents a line whose slope is equivalent to whole blood clearance in ml I/(sec × ml of renal tissue) of renal tissue. The y-intercept of this line represents fvv of renal parenchyma included in the ROI.

This method of determining regional blood flow and GFR is well described in several species using targeted ROIs and is the basis of the Patlak plot analysis.[10, 21-31]

Materials and Methods

Ten healthy adult (1–3 years old), domestic short hair cats housed for noneuthanasia research were used. The cats were normal based on physical exam, complete blood count, serum chemistry, urinalysis, and noninvasive blood pressure measurements. All cats had estimation of GFR using dynamic CT with Patlak plot analysis and SIC by the ICP/MS method.

All cats were administered IV acepromazine (0.05 mg/kg) and butorphanol (0.3 mg/kg) prior to anesthetic induction. This sedation protocol had the least effect on GFR in dogs as determined by nuclear scintigraphy.[32] The effect of this sedation protocol on GFR in cats is unknown. Anesthesia was induced with IV propofol (up to 4 mg/kg) at the lowest dose possible to achieve endotracheal intubation. The effect of this induction protocol is also unknown in cats. In dogs, no differences were found in GFR measurement obtained by dynamic CT among the induction agents propofol, thiopental, and etomidate on patients maintained with isoflurane.[33] Anesthesia was maintained with isoflurane and oxygen. Bradycardic or hypotensive cats received a total dose of 0.1 mg of atropine via IV injection during anesthesia. A ventilator was used for all cats at 12 breaths per minute. Lactated Ringer's solution was administered continuously at 5 ml/kg × h as maintenance for a minimally invasive anesthetic procedure. Hyperventilation was used prior to CT scanning to limit effects of respiratory motion.

Dynamic CT was performed using a third generation, single-slice helical scanner (GE CTI, GE Medical Systems™, Milwaukee, WI). The protocol consisted of a precontrast abdominal scan using 5-mm transverse helical slices obtained at 120 kVp, 200 mAs, a pitch of 1 and a tube rotation time of 1 s. Using images from the precontrast scan, a site for dynamic acquisition during and after the administration of iohexol was chosen. Slice location included both kidneys near each hilus and the aorta. A bolus infusion of iohexol (240 mgI/ml) (Omnipaque 240TM, Nycomed Inc., Princeton, NJ) at a dose of 300 mg iodine/kg was given via a cephalic catheter using a pressure injector (Medrad Visitron CT™, Pittsburgh, PA) at a rate of 3 ml/s. Just prior to initiation of the bolus injection, 10-mm-thick slices were acquired at the chosen location every fifth second for 115 s. All images were acquired using a 512 × 512 matrix. Following the dynamic study, both kidneys were scanned one final time using the same parameters described for the precontrast scan to determine their volume.

Images were transferred onto a personal computer and evaluated using an open-source imaging software program (Osirix™ v.3.6.1, Geneva, Switzerland). ROIs were drawn manually around each kidney, excluding the renal hilus and within the majority of the diameter of the abdominal aorta on each of the dynamic images and on a precontrast image obtained at the same scanning plane (Fig. 1). All regions were drawn at a window width and level of 400 and 40 HU by the same author (LAG). Attenuation data for each ROI were exported into a spreadsheet (Microsoft® Excel® for Mac 2011, Microsoft Corporation, Redmond, WA) for analysis. The mean attenuation value measured on the precontrast image for each kidney and the aorta was subtracted from the attenuation of each dynamic image to yield the corrected attenuation at these sites. TACs for each kidney and the aorta were generated from these data (Fig. 2). For each kidney, a Patlak plot was generated where math formula (math formula, corrected aortic HU value; math formula, area under the aortic TAC) was plotted against math formula (math formula, corrected kidney HU value; b(t), corrected aortic HU value) to yield a linear representation of GFR that has a slope equivalent to blood clearance of iodine in units of ml I/(sec × ml of renal tissue). Plasma clearance of iohexol (GFR) per kidney was then calculated by multiplying the slope of the line by 60 s/min to obtain results in ml I/min × ml, then (1-packed cell volume) to obtain plasma clearance rather than whole blood clearance, and, finally, the volume of each kidney to obtain results in ml I/min. The results were corrected for each cat's weight and reported in ml I/min × kg.

Figure 1.

The location for the transverse scan was chosen after obtaining a noncontrast abdominal study and matches the scout image (A). Regions of interest (ROIs) surrounding each kidney (white) and within the aorta (black) at four different times (B, 10 s; C, 15 s; D, 45 s; E, 75 s) of the dynamic computed tomography (CT) scan. Ten-millimeter-slice thickness was used and images were acquired every fifth second for 115 s. Right is to the left and dorsal is at the bottom of the images.

Figure 2.

Time attenuation curve (TAC). Attenuation in Hounsfield units as a function of time for the aorta and right and left kidney. TACs for all cats had a similar shape and the aortic curve was characterized by several recirculation peaks early in the scan. HU, Hounsfield units; and sec, seconds.

The iohexol injection used for the CTGFR study was also used for SIC determination based on the ICP/MS method. Three to four milliliters of blood were sampled via the jugular vein at 2 h, 3 h, and 4 h after the administration of iohexol. Blood was allowed to clot and serum was separated from the sample and submitted for analysis of iohexol concentration and GFR determination.

For each diagnostic method, the average GFR, its standard deviation and range were calculated. These descriptive statistics were intended to characterize variability in GFR reference values for each diagnostic method in normal patients. The diagnostic methods were compared and their differences (bias) estimated via Bland-Altman analysis [34] using a statistical linear mixed model that accounted for each individual cat as a blocking factor within which GFR was measured with each method. The model was fitted using the MIXED procedure of SAS statistical software (Version 9.2, SAS Institute Inc., Cary, NC). The correspondence between CTGFR and SIC was assessed using rank correlation analysis. The Spearman correlation coefficient between the two techniques was estimated and tested for evidence of a linear association in GFR between their relative rankings in the assessment of renal function. Pearson correlations between each method of GFR estimation and the lowest and highest blood pressures obtained for each cat during anesthesia were also estimated.


Aortic and renal TACs for all cats had a shape similar to those in Fig. 2. A sharp peak in aortic attenuation was seen at 5 or 10 s in all cats. Up to two additional smaller peaks were seen beyond this point in the aortic TAC. In the renal TACs, a small, broad initial peak was seen at 15 or 20 s in all cats. Following the initial peaks, there was a gradual decrease in aortic attenuation and a gradual increase in renal attenuation until the end of the 115-s acquisition time in all cats.

The initial peak in aortic and renal TACs are due to incomplete mixing creating inhomogeneous contrast medium distribution in blood (bolus) and a recirculation phenomenon.[22, 31] Data points on the TACs indicating this incomplete mixing and recirculation were not included in the construction of the Patlak plot. [22, 31] Exclusion of these recirculation peaks seen on the TAC from the graphical analysis removed outliers associated with Patlak plots and resulted in decreased variability (denoted by an increase in R2) in all but one cat. If a plot of the right, left, or total kidney density was characterized by high variability of R2 < 0.95 on the Patlak plot analysis, the images were analyzed for respiratory motion and images exhibiting motion were eliminated from Patlak plot graphical analysis. A Patlak plot of the right and left kidneys for the same cat in Figs. 1 and 2 is shown in Fig. 3.

Figure 3.

Patlak plot for cat seven in Table 1. The slope of each line, α/V, represents whole blood clearance in ml iodine/sec × ml renal tissue. The y-intercept represents fractional vascular volume (FVV).

Measurements of GFR using each method are summarized in Table 1. The average global CTGFR estimate was 1.84 ml I/min × kg (SD = 0.43; range = [1.22, 2.45]). The average GFR estimate determined by SIC was 2.46 ml I/min × kg (SD = 0.58; range = [1.72, 3.69]). The average CTGFR value was significantly less than the average GFR value determined SIC (P = 0.001). The difference (bias) between methods was significantly different from zero (P = 0.001) and consistently negative across cats with an estimated bias of −0.62 and an estimated standard error of 0.1307 (Fig. 4).

Table 1. Glomerular Filtration Rate Obtained for Each Cat Using Serum Iohexol Clearance (SIC) and Computed Tomography GFR (CTGFR). Individual and global GFR values are shown for the CTGFR method
GFR (ml I/min × kg)
  1. *Exceeds reference range of 1.15–2.73 ml I/min × kg provided by Michigan State University Diagnostic Center for Population and Health for SIC. GFR, glomerular filtration rate; CT, computed tomography; CTGFR computed tomography glomerular filtration rate; and SD, standard deviation.

Figure 4.

Bland-Altman analysis showing a bias of −0.62 (solid line) between glomerular filtration rate (GFR) estimations by computed tomography glomerular filtration rate (CTGFR) and serum iohexol clearance (SIC). This indicates consistent underestimation of the CTGFR method. CTGFR, computed tomography glomerular filtration rate; SIC, serum iohexol clearance; and GFR, glomerular filtration rate.

The Spearman rank correlation coefficient between the two techniques was 0.72 and was significantly different from zero (P = 0.023), providing evidence for a linear association in the relative rankings of GFR assessment between the two techniques. The scatterplot shows the relationship between the GFR values determined by each method (Fig. 5).

Figure 5.

Scatterplot showing the relationship between GFR values determined by dynamic CT and serum iohexol clearance (SIC). The Spearman rank correlation coefficient was estimated to be 0.72 (P = 0.0234).

Three of the cats were mildly hypotensive (≤80 mmHg) during the dynamic CT evaluation. Each was given atropine as described previously. Evidence did not support a linear association between CTGFR and highest blood pressure reading (P = 0.327), CTGFR and lowest blood pressure reading (P = 0.111), SIC and highest blood pressure reading (P = 0.198), or SIC and lowest blood pressure reading (P = 0.084).


In people, a correlation as high as 0.97 has been reported when comparing CTGFR with GFR determination by nuclear scintigraphy.[21] The correlation of global CTGFR with SIC in the current study, though statistically significant, is not as strong as global CTGFR correlations in people and pigs. [21, 22, 35] This may be due to our smaller sample size, differences in protocol, or may reflect true differences in results of GFR estimation by both methods. Recently, a procedure and reference range was published using CTGFR in dogs.[23] In an evaluation of CTGFR in dogs and pigs, the dynamic study was obtained by scanning every 2 and 4 s, respectively.[22, 23] We used a longer interscan delay, more closely matching human protocols.[10, 21, 27, 28] A longer interscan delay results in fewer images for construction of the Patlak plot, which may lead to a weaker correlation between CTGFR and SIC. However, the average final R2 value for all Patlak plot lines was 0.97 (SD = 0.02) for all cats.

The Patlak plot assumes a two-compartment model where the tracer leaves one compartment, crosses a permeable membrane, and irreversibly enters a second compartment. In any ROI within the kidney, there are cells, vasculature, interstitial space, and tubules. Cellular uptake of iohexol is negligible. The renal interstitial space is small.[35, 36] Additionally, there is free communication between the interstitial and vascular spaces, making them, effectively, part of the same compartment.[36] As such, iohexol within any ROI is contained in the vascular space and tubules. Iohexol leaves the vascular compartment irreversibly through the glomerular membrane and enters the tubules fulfilling pharmacokinetic properties for a two-compartment model and Patlak plot analysis.

To determine excretion rate of iohexol and estimate GFR, the contribution of attenuation from the vascular space and tubules must be separated from each other. Using CT, the attenuation of iohexol within the vascular compartment can be measured by including a major vessel, such as the aorta, in the scanning plane. Iohexol concentration within the aorta mimics that within renal vasculature. The ability to measure the contribution of vascular iohexol within an ROI allows for calculation of the amount of iohexol in the tubules. Additionally, iohexol within the tubules cannot leave the kidney during the analysis period, or the properties of a two-compartment model are no longer fulfilled. This is the basis of the scan time duration, 115 s.

The shape of the aortic and renal TACs was similar for each cat and similar to those previously reported in dogs and pigs.[22-24] After the recirculation peaks occurred, the aortic TAC had a mildly negative slope indicating decreasing iohexol concentration over time. The renal slope was mildly positive indicating increasing iohexol concentration over time. Recirculation peaks were not included in the analysis because the Patlak plot analysis requires complete mixing of contrast medium.

A significant difference was found between mean GFR values determined by SIC and CTGFR. A consistent underrepresentation (bias = −0.62) of GFR values was found based on dynamic CT. A bias of zero indicates that results of either method could be used interchangeably; however, given consistent underestimation, a set of reference ranges for CTGFR should be obtained using a higher sample number. The reason for the underrepresentation of GFR values determined by dynamic CT relative to SIC is not understood fully. Such an underestimation has been found previously.[21, 23] Effects of anesthesia, transient contrast-induced nephrotoxic effects, and hematocrit differences between small versus large vasculature (Faraeus effect) have all been theorized to result in underestimation of CTGFR values. [21-23] According to the Faraeus effect, hematocrit measurements are higher in large vessels than in small vessels. An overestimation of hematocrit due to sampling a large peripheral vein may result in an elevated GFR calculation when converting whole blood clearance to plasma clearance.

As discussed previously, the Patlak plot model assumes two compartments and the interstitial compartment is considered to be negligible and is, consequently, ignored. Initially, contrast medium in the vascular space will diffuse freely into the interstitial space.[36] Thus, there is a net flow of contrast medium from the vasculature to the interstitium. In the Patlak analysis, this interstitial flow of iohexol could contribute theoretically to an increase in the elimination constant of iohexol from the vascular space and, therefore, to a higher CTGFR value.[11, 37, 38] At some point in time, equilibrium between influx and efflux of iohexol between the interstitial space and vascular space occurs. After this equilibrium, the concentration of iohexol in the interstitial space exceeds that in the vascular space and there is net flow of iohexol from the interstitium back to plasma, potentially contributing to a falsely low CTGFR value. The backflow of iohexol during this time is considered small, if measurable at all.[37] The slope of the Patlak plot will equal the GFR only at the equilibrium point.[35, 37] In people, the equilibrium point was later than 97 s.[37] Patlak plot measurements must be obtained before iohexol is present in the ureters, which usually occurs within 2 min. These time constraints leave a narrow window to begin the scan at the equilibrium point and achieve a significant number of data points necessary for analysis before contrast medium enters the ureters. Theoretically, the time period used for data acquisition in this study should result in falsely elevated CTGFR values as most data points were likely prior to the equilibrium point; however, this seems to have an insignificant effect on results in animals with a normal volume of interstitial tissue.

The volume of interstitial space in feline kidneys is unknown. In people, the interstitial space in clinically normal kidneys is about 8.4% in the cortex and 17.6% in the outer medulla. In patients with acute renal failure, the interstitial space increased by about 10% in the cortex and medulla. [38] Overestimation of CTGFR values can occur when there is increased interstitial space.[35, 37] Inclusion of the interstitial space as a third compartment for pharmacokinetic GFR analysis resulted in consistent overestimation of values in people.[37] In people, it has been suggested that measurement of GFR using dynamic CT should be reserved for patients with chronic rather than acute kidney disease.[35]

The GFR determination by SIC exceeded the reported reference range in three of the cats. Interestingly, these cats had the three highest GFR measured by Patlak plot analysis. Only one of these cats received atropine during the CT examination. The reason for the increased measured GFR in these cats is unknown. An early manifestation of thyroid disease or other causes of hypertension is possible; however, these cats were young and otherwise healthy. The highest and lowest noninvasive blood pressure values obtained during anesthesia had no correlation with GFR results obtained from either method, which is not surprising given the renal ability to autoregulate blood flow within the blood pressure ranges recorded during the CTGFR procedure.

An important limitation of this technique is that estimation of GFR is calculated using a single 10-mm-slice thickness of kidney. A risk of misrepresenting GFR in kidneys with nonuniform disease exists.[22, 39] The final postcontrast scan is essential to evaluate morphologic abnormalities that indicate potential for misrepresenting GFR and for accurately calculating the volume of functional renal tissue. Also, as previously noted, the renal interstitial compartment is ignored which may result in inaccuracy. An additional limitation of this technique is the necessity of anesthesia to limit motion. Although no correlation was seen between blood pressure measurements and CTGFR values, the effect of anesthesia and sedation on this procedure is unknown.

An advantage of CTGFR is that GFR can be calculated with minimal alteration to standard pre and postcontrast abdominal scans. Results are relatively rapid once a spreadsheet is designed for calculations. Morphologic and functional information about the kidneys can be obtained using a single modality. Also, individual kidney function can be assessed, which may be important prior to nephrectomy/nephrotomy or donor/recipient transplant patients. Finally, a dose of iohexol that is one-fourth of that is used for GFR determined by iohexol clearance correlated with traditional methods of GFR testing in people and pigs, reducing risk associated with contrast-induced nephrotoxicity.[21, 22] In this study, the iohexol dose was dictated by the protocol for SIC.

In conclusion, the average global CTGFR estimate was 1.84 ml/min × kg (SD = 0.43; range = [1.22, 2.45]). CTGFR consistently underestimated renal function relative to GFR determined by iohexol clearance. Regardless of this bias, the positive correlation between CTGFR and iohexol clearance indicates potential for the clinical usefulness of this testing method. Research on clinically affected cats is needed to further validate dynamic CT and Patlak plot analysis as a tool for GFR estimation.