• Open Access

Glomerular Filtration Rate Estimated by 3-Sample Plasma Clearance of Iohexol in 118 Healthy Dogs


  • Previously presented in part at the 16th Annual Congress of the European College of Veterinary Internal Medicine Companion Animals (ECVIM-CA), Amsterdam, The Netherlands, September 14–16, 2006.

Corresponding author: Nicholas H. Bexfield, BVetMed, DSAM, DipECVIM-CA, MRCVS, Department of Veterinary Medicine, University of Cambridge, Cambridge, CB3 0ES, UK; e-mail: nb289@cam.ac.uk.


Background: Glomerular filtration rate (GFR) decreases in the aging human kidney, but limited data exist in dogs.

Hypothesis: There is an effect of age and body size on estimated GFR in healthy dogs.

Animals: One hundred and eighteen healthy dogs of various breeds, ages, and body weights presenting to 3 referral centers.

Methods: GFR was estimated in clinically healthy dogs between 1 and 14 years of age. GFR was estimated from the plasma clearance of iohexol, by a compartmental model and an empirical correction formula, normalized to body weight in kilograms or liters of extracellular fluid volume (ECFV). For data analysis, dogs were divided into body weight quartiles 1.8–12.4, 13.2–25.5, 25.7–31.6, and 32.0–70.3 kg.

Results: In the complete data set, there was no trend toward lower estimated GFR/kg or GFR/ECFV with increasing age. GFR decreased with age in dogs in the smallest weight quartile only. A significant negative linear relationship was detected between body weight and estimated GFR/kg and GFR/ECFV. Reference ranges in different weight quartiles were 1.54–4.25, 1.29–3.50, 0.95–3.36, and 1.12–3.39 mL/min/kg, respectively. Standardization to ECFV rather than kilogram body weight did not produce substantial changes in the relationships between GFR estimates and age or weight.

Conclusions and Clinical Importance: Interpretation of GFR results for early diagnosis of renal failure should take into account the weight and the age of the patient for small dogs.

The indications for accurate assessment of renal function in veterinary medicine are widespread.1 Examples include screening for the presence of renal disease in animals with nonazotemic polyuriaa or mild increases in plasma creatinine concentration, screening in breeds with a predisposition to familial renal disease, presurgical or posttreatment monitoring, dosage guidance when using drugs excreted by the kidney2 and evaluating effects of dietary or other therapeutic interventions on renal function.3,4 Sequential measurements of glomerular filtration rate (GFR) also provide useful information on alterations in renal function over time.

GFR is regarded as the best overall index of renal function in health and disease, and is superior to plasma creatinine or cystatin C concentrations and other simpler tests of renal function.5 GFR cannot be measured directly but is estimated using the clearance of a filtration marker.1 Various exogenous and endogenous markers have been used. Of these, the urinary clearance of the fructose polymer inulin has long been regarded as the reference method for determining GFR in humans and dogs.

An alternative to urine collection procedures is to determine plasma clearance of a marker, such as inulin, iodine-containing radiographic contrast media, radionuclides or creatinine.6–9 The expense, laboratory analysis, and lack of availability generally preclude the use of inulin, whereas the use of radionuclides requires access to a nuclear medicine facility. The radiographic contrast agent iohexol has been used extensively in human nephrology.10–12 Estimation of GFR by the plasma clearance of iohexol in dogs has been reported,8,13–21 and has shown good correlation with methods involving radionuclide-labeled markers and exogenous creatinine clearance.8,15

Limited sampling strategies for plasma clearance procedures are used in human and veterinary medicine for a simple and clinically acceptable estimation of GFR. Correction formulas must be applied when these strategies are used, with the most widely employed being the Brøchner-Mortensen (BM) formula.22 Some veterinary publications make use of this formula for corrections,14,16,18,21 but species differences have not been systematically evaluated. In this study, when predicting the total clearance from the 1-compartmental clearance, we used a species-specific correction formula calculated from data in dogs with normal or moderately decreased renal function.18

Morphological changes are noted in the human glomerulus with aging and there is an associated decrease in GFR.23–25 Estimated GFR in healthy humans over the age of 60 years is 20–30% lower than corresponding GFR in individuals under the age of 50 years.23 Although similar age-related histological changes are documented in dogs,26,27 limited data confirm a corresponding decrease in GFR with age.b GFR remained unaltered in geriatric dogs > 4 years after uninephrectomy.28 In another study, GFR was higher in puppies than in 6 to 9-year-old dogs, although an effect of age was not apparent when GFR estimates were normalized to liters extracellular fluid volume (clearance/ECFV) rather than to body weight (clearance/kg).20 Although not extensively studied in veterinary medicine, some published data indicate differences in GFR among mammalian species of different body sizes29 as well as among dogs of different breeds and body weights.c,30,31

The aim of this prospective study was to estimate GFR in healthy dogs of different ages and body sizes by a 3-sample method for plasma clearance of iohexol.

Materials and Methods


The study was conducted at 3 centers (University of Cambridge, Cambridge, UK; Norwegian School of Veterinary Science, Oslo, Norway; and DeKompaan Veterinary Clinic, Ommen, Holland, the Netherlands) during 2002–2005 with the majority of clearance studies performed during 2005. Ethical approval and written consent were obtained from the owners according to the appropriate local and national regulations and ethical committees. The study population consisted of clinically healthy staff- and client-owned dogs. Dogs were undergoing neutering or minor dental treatment, or were owner-volunteered and included 54 dogs that were regular blood donors. Blood donors were owner-volunteered pet dogs that donated blood a maximum of twice a year, and no dog had donated blood for more than 2 years. The mean dog age was 6.7 years (range, 1–14 years) with a mean body weight of 12.8 kg (range, 1.8–70.3 kg). Labrador Retriever (n = 14), cross breed (n = 14), Golden Retriever (n = 6), Greyhound (n = 5), and Boxer (n = 5) were the most common breeds, but 49 breeds were represented. All dogs were considered clinically healthy by the owners at the time of the investigation, and this observation was supported by findings on physical examination. Complete CBC, serum biochemistry (20 variables), and urinalysis (voided urine) were performed to further assess health status. Exclusion criteria were a previous history of renal or urinary tract disease or abnormal clinicopathological findings. Animals with more than trace proteinuria by dipstick analysis were excluded. If other dipstick abnormalities were detected, urine sediment analysis was performed. If abnormalities were identified on sediment analysis, these animals also were excluded. One hundred and eighteen dogs were included in the study.

Clearance Procedure

A standard test protocol was followed at all centers. Food was withheld for 12 hours before the procedure, whereas free access to drinking water was allowed throughout the study. Iohexold was administered at a dosage of 60–300 mg I/kg as a bolus over approximately 30 seconds via a cephalic catheter followed by a heparinized saline flush. The dose of iohexol was measured from the syringe volume. The completion of the injection represented time-zero. Five milliliters of blood was collected from the jugular vein into a lithium heparin tube before and at approximately 2, 3, and 4 hours after injection. The exact time of sampling was used in the calculations. Blood samples were centrifuged at 1,500 × g for 10 minutes; plasma was frozen at −20 °C until analysis.

Laboratory Analysis of Iohexol

The plasma standards (1, 10, 50, 100, and 250 μg/mL) were made by diluting the stock solution (647 g/L, corresponding to 300 g I/L) of iohexold with pooled plasma obtained from 10 healthy, fasted dogs.

Samples and standards were deproteinized by adding 1 volume of acetonitrile/ethanol (1:1, v/v), left overnight, and then centrifuged at 15,000 × g for 30 minutes. The supernatant was diluted by adding 3 volumes of high-performance liquid chromatography (HPLC)-grade watere before injection. All steps were carried out at 4 °C.

Adjustment of a previously published method for iohexol analysis was performed.32 Iohexol in plasma from dogs was determined by HPLCf consisting of a quaternary pump, an online degasser, an autosampler, and a diode array detector with a 13 μL flow cell and 10 mm path length.

The analyses were performed on a Waters Spherisorb, 5 μm ODS2, analytical reversed-phase column (4.6 × 250 mm ID).g The mobile phase was 5% acetonitrile in water with a flow rate of 0.9 mL/min. Total run time was 60 minutes and the injection volume was 15 μL for all samples. Column temperature was 25 °C.

The iohexol concentration was calculated from the height of the exo-iohexol absorbance peak at 244 nm, using a 4 nm slit and a peak width of >0.2 minutes (4 seconds). The retention times for the 2 stereoisomers were 7.5 and 8.6 minutes, respectively. The accuracy was calculated from the following equation:


The results from 5 parallel standards in plasma were as follows: 5.5% at 1 μg/mL, −2.5% at 10 μg/mL, and 0.3% at 100 μg/mL.

Iohexol analysis was performed at the Norwegian School of Veterinary Scienceh except in 30 dogs where analysis was performed at The South West Thames Institute for Renal Research Laboratoryi by a very similar method.

Pharmacokinetic Analysis


where D is the dosage and AUC is the area under the curve. AUC was calculated by a commercially available pharmacokinetics computer programj by use of a 1-compartmental model based on the terminal monoexponential slope of the curve as defined by samples collected at 2, 3, and 4 hours.

A calibration procedure by use of the empirical polynomial regression formula18:


was performed to predict the complete-curve plasma clearance, which is based on 10 samples over 6 hours (CL2comp). The clearance value then was standardized to kilogram body weight.

Similarly, a calibration procedure by use of the empirical regression formula:


was performed to predict the complete-curve GFR/ECFV, ie, total plasma clearance of iohexol standardized to liters of ECFV, based on 10 samples over 6 hours (GFR/ECFV2comp), by use of the terminal elimination rate constant K10 defined by samples taken at 2, 3, and 4 hours (GFR/ECFV1comp).18

GFR estimates predicted by use of the dog-specific correction formula from the 3-sample values18 also were compared with the estimates predicted by use of the BM formula (Cl = 0.990778 Cl1–0.001218 Cl12).22

Statistical Analysis

Continuously distributed variables were expressed as mean values with 95% confidence intervals calculated by Student's procedure.33

The standard deviation (SD) and total ranges are given as indexes for dispersion. Reference ranges are expressed as mean ± 2SD. The 95% confidence intervals for the upper and lower boundaries in the reference ranges were calculated by a previously described procedure.34

All tests in the analyses were performed in a 2-tailed manner with a significance level of 5%.

Linear regression analysis35 was used to study the influence of each independent variable (age and body weight) on the 2 dependent variables (GFR/kg and GFR/ECFV).

The simultaneous influence of age and body weight including interaction was performed by multiple regression analysis.35 The individuals were grouped in body weight quartiles. These groups were compared for GFR and GFR/ECFV by the Student-Newman-Keuls procedure in an analysis of variance (ANOVA).35


No adverse clinical signs were observed during or after the infusion of iohexol.

There was no trend toward lower estimated GFR/kg or GFR/ECFV with age (Fig 1A and B), and age alone was found to explain <1% of the variation in both GFR/kg and GFR/ECFV.

Figure 1.

 Relationship between estimated glomerular filtration rate (GFR)/kg body weight (A; 1-compartment corrected to 2-compartment plasma clearance of iohexol) or GFR per liter extracellular fluid volume (B; 1-compartment corrected to 2-compartment plasma clearance of iohexol) and age in 118 healthy dogs.

A significant negative linear relationship was detected between body weight and estimated GFR/kg (P < .01) and GFR/ECFV (P < .01) (Figs 2A and B). Body weight explained 15.6% (R2= .156) and 30.9% (R2= .309) of the variation in estimated GFR/kg and GFR/ECFV, respectively.

Figure 2.

 Relationship between estimated glomerular filtration rate (GFR)/kg body weight (A; 1-compartment corrected to 2-compartment plasma clearance of iohexol) or GFR per liter extracellular fluid volume (B; 1-compartment corrected to 2-compartment plasma clearance of iohexol) and kg body weight in 118 healthy dogs.

No significant interaction was found between age and body weight with regard to GFR/kg and GFR/ECFV. However, body weight and age nested for weight quartiles was found to explain 18.3 and 31.7% of the variation in estimated GFR/kg and GFR/ECFV, respectively. The smallest body weight quartile was the only quartile that contributed significantly in the model.

A significant difference between body weight quartiles (P≤ .001) was detected for both GFR/kg and GFR/ECFV (Table 1). Both variables were significantly different in the smallest quartile compared with the others (P≤ .003), as reflected in the reference ranges (Table 2).

Table 1.   Means (X), standard deviation (SD), and confidence intervals (CI) for the estimates of GFR and GFR/ECFV based on plasma clearance of iohexol in 118 healthy dogs, within weight quartiles of kg body weight.
Weight QuartileGFR (mL/min/kg)GFR/ECFV (mL/min/L)
  1. GFR; glomerular filtration rate; ECFV, extracellular fluid volume.

1: 1.8–12.4 kg (n = 29)2.89 (0.68)2.64–3.1516.27 (3.40)14.97–17.56
2: 13.2–25.5 kg (n = 30)2.40 (0.55)2.19–2.6013.80 (2.69)12.77–14.80
3: 25.7–31.6 (n = 30)2.16 (0.60)1.93–2.3812.38 (3.34)11.14–13.63
4: 32–70.3 kg (n = 29)2.25 (0.57)2.04–2.4711.54 (2.84)10.37–12.53
Table 2.   Reference ranges for estimated GFR in small and large dogs as corrected plasma clearance of iohexol, and estimated GFR/ECFV as corrected plasma disappearance rate constant for iohexol, and the 95% confidence intervals for the upper and lower limit of the reference range, respectively.
Weight QuartileReference Range95% Confidence Interval for the
Low Reference Range Value
95% Confidence Interval for the
High Reference Range Value
Lower LimitUpper LimitLower LimitUpper Limit
  1. GFR; glomerular filtration rate; ECFV, extracellular fluid volume.

a. Clearance (mL/min/kg)
1: 1.8–12.4 kg (n = 29)1.54–
2: 13.2–25.5 kg (n = 30)1.29–3.501.031.742.794.71
3: 25.7–31.6 kg (n = 30)0.95–3.360.761.292.684.52
4: 32–70.3 kg (n = 29)1.12–3.390.891.522.694.58
b. GFR/ECFV (mL/min/L)
1: 1.8–12.4 kg (n = 29)9.46–23.077.5112.8018.3031.21
2: 13.2–25.5 kg (n = 30)8.42–19.196.7111.3315.2825.83
3: 25.7–31.6 kg (n = 30)5.70–19.074.547.6715.1925.67
4: 32–70.3 kg (n = 29)5.78–17.024.587.8113.5823.16

GFR/ECFV in the second smallest quartile was significantly larger compared with the largest quartile (P= .004). No significant differences in GFR/kg were detected among the 3 highest quartiles.

No significant linear correlation was detected between age and weight (r= .07, P= .42). However, a significant effect of age on GFR/kg (P= .05) was detected in the lowest weight quartile (Fig 3). In this case, age alone explained 12.5% of the variation in GFR. No similar pattern was detected for GFR/ECFV in the other quartiles. A linear relationship between estimated GFR/kg and GFR/ECFV was observed, but values from individual dogs displayed substantial scatter from the regression line (Fig 4).

Figure 3.

 Decrease in estimated glomerular filtration rate/kg body weight (1-compartment corrected to 2-compartment plasma clearance of iohexol) with increasing age in healthy dogs in each of the weight quartiles (A; 1.8–12.4 kg, B; 13.2–25.5 kg, C; 25.7–31.6 kg, D; 32–70.3 kg).

Figure 4.

 Relationship between estimated glomerular filtration rate (GFR)/kg body weight (1 compartment corrected to 2 compartment plasma clearance of iohexol) and GFR/extracellular fluid volume (terminal rate constant of the plasma decay curve of iohexol, corrected to 2-compartment estimate) in 118 healthy dogs.

In most dogs, GFR estimates predicted by use of the dog-specific correction formula from the 3-sample values18 were closely related to the estimates predicted by use of the BM formula.22 For approximately 80% of dogs, the GFR estimates by use of the BM formula were 0–5% higher than by use of the dog-specific formula. However, at high clearance values (ie in the smallest dogs), the predicted values by the BM formula were 60–100% higher (Fig 5).

Figure 5.

 Relationship between glomerular filtration rate (GFR) estimates by plasma clearance of iohexol (CL/min/kg) as predicted from a dog-specific correction formula18 and GFR estimates predicted by use of the Brøchner-Mortensen (BM) formula from human medicine.22


The results of this study document an effect of body size on estimated GFR, which was comparatively larger than the effect of age. GFR estimated by plasma clearance of iohexol was of similar magnitude to values published previously.8,13–15,17,19,20 Estimated GFR was found to be higher in dogs in the smallest weight quartile. A previous study examining plasma clearance of iohexol in 31 healthy dogs similarly discovered decreasing GFR or GFR/ECFV with increasing body size.17 The estimated GFR values observed in dogs in the other weight quartiles displayed a wide reference range. Although this effect also was observed in 34 healthy dogs between 8 and 54 kg,31 the size effect in our dogs weighing between 1.8 and 70.3 kg was more pronounced. This finding is in accordance with recent data in which breed differences in estimated GFR in healthy dogs were observed.c In the previously reported study, too few dogs of each breed were available to evaluate a breed effect. In 93 dogs with normal and decreased renal function, a change in dosage of inulin was introduced after the observation that clearance was strongly associated with body weight with the smallest dogs demonstrating rapid clearance of inulin.30

The weak trend of age, which was observed only in the smallest weight quartile, and also not in the 14 Labrador Retrievers, did not support our original hypothesis, based on results from human medicine, that GFR would decrease with age.23–25 There are several possible explanations for this finding. Because of the effect of body size on estimated GFR and the wide range of body sizes in both young and old dogs in this study, a larger sample size may have been required to document an age effect. Alternatively, because of the shorter life span of dogs compared with humans, decreases in estimated GFR may not occur to the same extent, despite similar histopathological changes. The increased longevity of small versus large breeds does not explain the lack of age effect in this study, because there was no correlation between body size and age.

There was little overlap in GFR values in the smallest dogs and dogs of intermediate and large sizes, although some very small dogs had clearance values similar to dogs of other sizes. Evaluation of renal function by use of plasma disappearance or renal clearance must be done with caution in dogs of extreme body sizes. These observations most likely also influence the wide confidence intervals for the reference ranges in each weight quartile. A wide reference range is most likely a biological effect related to GFR itself, not specifically related to the use of a simplified plasma clearance method.29,31 Similarly wide reference ranges also are observed in human medicine. A reference range for estimated GFR can be expected to be wide because GFR is known to be influenced by several nonrenal factors such as hydration status, protein intake, sodium balance, and circadian rhythm (day-to-day) variation for the individual dog.1,7,36 Dogs were fasted overnight and allowed free access to water to decrease methodological variability in the present study.

Unfortunately, because of reduced capacity of the initial laboratory, analysis of iohexol was performed at a 2nd laboratory that used very similar methodology. Although statistical analysis was not performed to compare the age, weight, sex, and breed between populations studied in the 2 laboratories, samples were randomly allocated to each laboratory. It is possible, however, that the results may have been influenced by this factor.

The wide reference range warrants caution when interpreting individual results to detect early renal disease. Although animals with estimated GFR below the reference range by definition have abnormal renal function, it is possible that some animals also have abnormal renal function despite GFR results within the reference range. In this large and heterogenous population of client-owned animals, it is not surprising that wide reference ranges were encountered, and wide reference ranges are similarly observed in human nephrology. These reference ranges should be applicable to the population of animals seen in clinical practice, however.

Indexing to ECFV rather than kg body weight did not produce substantial changes in the relationships between GFR estimates and age or weight. There are several physiological arguments for the use of standardization to liters of ECFV rather than kg body weight.37,38 When comparing estimated GFR values in puppies and adults, the differences between groups disappeared by standardization to ECFV.20

Estimation of GFR by the plasma clearance of iohexol is attractive for clinical practice and research. Advantages of iohexol over other markers include its lack of radiation hazard, wide availability, and the stability of iohexol in plasma, facilitating sampling and shipment of unfrozen samples to remote laboratories.

Plasma clearance methods or limited-sample strategies need appropriate validation.1,6,18 If its limitations are acknowledged, we consider a 3-sample iohexol plasma clearance method valid for clinical and research purposes.

Calculations are based on the plasma disappearance curve of the marker, which typically shows initial rapid reduction in plasma concentration (“distribution phase”) and a terminal monoexponential slope (“elimination phase”). The magnitude of the systematic error by use of a limited-sample approach differs with renal function. Although this error may be ignored when renal function is severely decreased, it typically represents 20–30% of the clearance value in healthy dogs.8,18,39 In a healthy dog, the initial distribution phase for iohexol typically lasts for 30–60 minutes, and virtually all of the iohexol is excreted after 6–8 hours. Some veterinary publications describe samples in both phases, necessitating collection of at least 4 samples over several hours.9,40 We consider it more convenient to use the terminal monoexponential slope,18 with sampling over 2 hours 2–4 hours after injection. A recent study in human medicine demonstrated no loss in accuracy by one of these approaches over the other.12

Although the accuracy of a 2-sample method was equal to that of a 3-sample method in a previous study,18 3 samples were used in this study to decrease the chance of errors and increase the accuracy of the GFR/ECFV estimates. Extended sampling times may be necessary in cases of severely decreased renal function, when excretion of a drug eliminated only by glomerular filtration is delayed.11 However, the protocol used in this study may be considered adequate in cases in which renal function is normal or moderately decreased, which is the case for most patients where estimation of GFR is indicated. In cases of severely decreased renal function (ie, markedly azotemic animals), use of full curve estimates of extended sampling times is likely more accurate.

The validity of the correction formula is an obvious concern when using limited-sample strategies. We used a dog-specific formula for calculation of the reference ranges (Table 2). The BM formula for correction from the monoexponential slope has been used for decades and is shown to be remarkably robust in human medicine.12,41 Because it has been applied in veterinary publications,14,16,18,21 we also studied the differences between GFR estimates by the BM formula and the dog-specific formula. The few dogs with severe discrepancies in calculated values (Fig 5) were very small dogs (2–6 kg body weight). This finding appears to support the hypothesis that the BM formula, which is based on a population of healthy adult humans, is reasonably accurate in large but not very small dogs.

No adverse drug reactions were observed in this study and in other studies with iohexol for GFR estimation in dogs.8,13–21 With the extensive and increasing use of radiographic contrast media for diagnostic imaging in all areas of medicine, the prospect of adverse drug reactions is a concern. Although data on adverse drug reactions in animals are scarce, adverse reactions were uncommon in a large prospective study of more than 300,000 human patients receiving radiographic contrast media, with a calculated prevalence rate of severe adverse effects of 0.04%.42 High doses of iohexol for radiographic contrast studies also appear quite safe even in very high-risk patients with markedly decreased renal function, although transient adverse drug reactions have been observed.12,43 A number of studies in human medicine have addressed the effect of radiographic contrast media on renal function, as summarized in a recent review.11 Although minor and transient adverse drug reactions were observed in up to 3% of patients in a large prospective study investigating immediate and delayed adverse drug reactions after use of iohexol, not a single severe or life-threatening adverse drug reaction was observed.44 Also, severe adverse drug reactions have been described with the use of inulin.45,46

This 3-sample method provides a reliable and convenient method for estimating GFR in dogs with normal or nearly normal GFR. When interpreting estimated GFR values, body size and the age of the patient (for small dogs) should be taken into account.


aArons J, Van der Heyden S, Lefebvre H, et al. Polyuria-polydipsia in dogs: usefulness of GFR assessment through creatinine clearance testing for differential diagnosis. J Vet Intern Med 2005;19:935 (abstract)

bQueau Y, Biourge V, Germain C, Braun JP, Watson AJD. Effect of aging on plasma exogenous creatinine clearance in dogs. J Vet Intern Med 2007; 21: 598 (abstract)

cLefebvre HP, Craig AJ, Braun JP. GFR in the dog: Breed effect. Proceedings of the 16th ECVIM-CA congress Amsterdam, 2006, 61 pp

dOmnipaque, Amersham Health, GE Healthcare, Oslo, Norway

eBarnstead, NANOpure ultrapure water system, Barnstead International, Dubuque, IA

fAgilent system (series 1100), Agilent Technologies Deutschland GmbH, Waldbronn, Germany

gWaters Corporation, Milford, MA

hNorges Veterinærhøgskole, Oslo, Norway

iSouth West Thames Institute for Renal Research Laboratory, St Helier Hospital, Surrey, UK

jWinNonlin 5.1, Scientific Consulting Inc, Apex, NC


The authors thank Petter Balke Jacobsen for input when establishing the iohexol analysis, The South West Thames Institute for Renal Research Laboratory, and Stig Larsen for help with statistical analysis. The authors are also grateful to the owners and patients for their participation in this study.

This work was done at The Department of Veterinary Medicine, University of Cambridge, Cambridge, UK; The Department of Companion Animal Clinical Sciences, Norwegian School of Veterinary Science, Oslo, Norway and R.J. deKompaan Veterinary Clinic, Ommen, Holland, the Netherlands. Grant: This work was supported by a grant from the Norwegian Research Council.