Abstract
Several publications suggest a potential association between the administration of Gadolinium-based contrast agents (GBCAs) and the onset of a rare but serious disease, Nephrogenic Systemic Fibrosis (NSF). The aim of this study was to determine the elimination time-course of Gadolinium (Gd) from skin tissue after application of GBCAs in rats. Seven different marketed GBCAs were injected on five consecutive days at a dose of 2.5Â mmol/kg bodyweight into the tail vein of Han-Wistar rats and the Gd concentrations were determined by Inductively Coupled Plasma Mass Spectrometry (ICP-MS) in skin biopsies taken at various time-points up to a year after the last injection. Most of the administered Gd was eliminated from the skin within a time-period of about 2 months. However, the repeated administration of linear GBCAs resulted in long-term retention of a small portion of the administered Gd in the skin tissue of rats, with substantially higher values observed in animals treated with non-ionic linear agents than in those that received ionic linear GBCAs. Following treatment with macrocyclic GBCAs, Gd values in the skin were in the same range as observed in the controls from day 24 post-injection onwards. In summary, we observed a correlation between the complex stability of GBCAs and the amount of residual Gd in the skin up to a year after application of GBCAs.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Introduction
Gadolinium-based contrast agents (GBCAs) used for Magnetic Resonance Imaging (MRI) contain a single paramagnetic Gd3+ ion tightly bound by a complexing agent. On the basis of the chemical structure, GBCAs can be divided into two distinct classes: the linear, open-chain agents and the macrocyclic agents [1]. Each class can be further divided into ionic and non-ionic agents. The different classes, as well as the different agents within each class, have different complex stabilities, i.e. propensities to retain the Gd3+ ion in the complex (Table 1).
The stability of GBCAs is characterized by two parameters, the kinetic inertia and the thermodynamic stability. Macrocyclic GBCAs are kinetically inert to Gd dissociation compared to linear GBCAs, as they possess a substantially higher kinetic stability, characterized by the dissociation half-life [2, 3]. As the estimated dissociation half-life of macrocyclic GBCAs is over 1,000Â years at pH 7.4, the thermodynamic stability of these GBCAs, characterized by log Ktherm and log Kcond, is of minor relevance [3].
The linear GBCAs possess a comparatively lower kinetic stability. For the stability of these GBCAs, the thermodynamic stability is a more relevant factor. The non-ionic GBCAs, such as Omniscan and OptiMARK, have a substantially lower thermodynamic stability compared with the ionic GBCAs, such as Magnevist or MultiHance [3].
Several publications have suggested a potential association between the administration of GBCAs and a relatively new disease, Nephrogenic Systemic Fibrosis (NSF) [4, 5]. In articles based on preclinical studies we have conducted, we have reported a correlation between the Gd concentration in skin tissue and the occurrence of NSF-like skin lesions in rats [6, 8]. In all such studies, we observed marked differences regarding the Gd concentration in various tissues of the animals after treatment with the different GBCAs. The most pronounced differences were observed in the skin. The highest Gd concentrations in skin tissue were observed after treatment with non-ionic linear GBCAs (Omniscan and OptiMARK). The lowest concentrations were observed after treatment with macrocyclic GBCAs (Gadovist and Dotarem) [7]. It is important to note that the measurement method (ICP-MS) used cannot distinguish between chelated and unchelated Gd. This fact led to the hypothesis that at least part of the Gd measured 5 days post-injection (p.i.) may still have consisted of chelated Gd, which may potentially be eliminated from the body at a later time-point. Therefore, the aim of this study was to determine the elimination time-course of Gd in skin tissue and the potential long-term retention of Gd in the skin after application of different marketed GBCAs.
Materials and methods
Animals
Han-Wistar rats (CRL:WI [GIx/BRL/HAN]IGS BR) were obtained from Charles River (Sulzfeld, Germany), each weighing 210–346 g. The animals were housed with a 12 h day/night rhythm and were given water and standard diet (ssniff R/M-H from Sniff, Germany) ad libitum. Animals were treated according to the German law on the protection of animals and with permission of the state animal welfare committee.
Study design
Six animals for each treatment group were selected at random and were treated in three experimental sets. Omniscan (set 1), Magnevist (set 1), Gadovist (set 1), MultiHance (set 2), ProHance (set 2), Dotarem (set 2), and OptiMARK (set 3) and saline (set 3) were injected into a tail vein once daily on five consecutive days at a dose of 2.5 mmolGd/kg bodyweight (b.w.). An untreated group (set 1) was included as a second negative control group. Skin biopsies were taken from the backs of the animals at various time-points up to 364 days p.i. (please refer for the different time-points to Fig. 1) under 3.5–4% isofluran inhalation with a biopsy punch (Biopunch 4 mm, Stiefel Laboratorium GmbH, Offenbach Germany) and the wounds were sutured with Vicryl rapide (Ethicon GmbH, Norderstedt Germany).
Gadolinium concentration in the skin biopsies taken from animals treated with Omniscan (red diamonds), OptiMARK (orange triangles), Magnevist (violet dots), ProHance (green circles) Dotarem (green triangles), Gadovist (green squares) and control animals (saline and untreated). The animals were injected on five consecutive days with 2.5Â mmolGd/kg body weight (b.w.) per injection and the Gd concentration in the skin is given as nmolGd/g skin
Pharmacokinetics of GBCAs in rats
To assess the pharmacokinetic parameters of GBCAs, Omniscan, Magnevist, MultiHance and Gadovist were injected once at a dose of 2.5 mmol/kg b.w. into the tail vein, and blood samples were taken from the cannulated carotid vein at various time-points p.i. (1, 3, 5, 10, 15, 30, 60, 90, 120, 180 and 240 min). The pharmacokinetic parameters were calculated with the help of WinNonlin (Pharsight Corporation, Montain View, Ca, USA), two-compartment model, g = 1/y2.
Gadolinium measurements
The Gd concentrations in the skin samples and serum were determined by Inductively Coupled Plasma Mass Spectrometry (ICP-MS 7500a, Agilent, Waldbronn, Germany) by measuring the most abundant isotope 158Gd. Before drying at 90°C, the skin samples (7-22 mg) were mixed with 50 µl of 1 µM Tb(NO3)3, serving as an internal standard. Afterwards 50 µl concentrated HNO3 and 20 µl H2O2 were added and heated to dissolve the tissue. The solution was filled up to 1 ml with H2O. The lower limit of quantification of the method is about 1 nmol/l or 0.05 nmol Gd/g skin [20, 21]. It should be noted that this method cannot distinguish between chelated and unchelated Gd.
Statistical evaluation
In order to calculate the areas under the curve (AUC), an exponential function was fitted to the data points by the help of the solver function of Excel (Microsoft Cooperation, Redmond, Wa, USA). Based in this function, the daily Gd exposure was calculated and summed up to the AUC for the investigated period. The AUCs were used to compare the treatment groups for which a control group (either saline or untreated) was available, i.e. for set 1 and set 3. The AUCs of the treatment groups were adjusted by using the difference in AUCs of the treatment group and the respective control.
Treatment differences were tested by a distribution-free Kruskal-Wallis one-way analysis of variance on ranks. Overall differences of the medians were tested by the Kruskal-Wallis test, whereas pairwise differences were tested by Dunn’s test. This distribution-free multiple comparison test does not need the assumption of normality and tests pairs of medians following the Kruskal-Wallis test. The significance level is adjusted on the basis of comparisons to keep the two-sided significance level of 5% for each experiment. Instead of means, this test uses average ranks. Calculations were performed using NCSS 2007 (NCSS, Kaysville, Ut, USA).
Results
For several weeks after the last injection, the Gd concentration in the skin decreased substantially in all treatment groups. About 60 days after treatment with linear GBCAs, a plateau phase for Gd concentration was observed. Following this phase, the Gd level decreased only marginally during the next 304 days (Fig. 1). After treatment with the macrocyclic GBCAs, the Gd values in skin tissue returned to the same range as that observed in the control animals from about day 24 post-treatment onwards. It should be noted that we could detect small amounts of Gd in the skin of control animals, most obvious in the skin of the untreated animals (set 1). No macroscopic skin changes were observed in any treatment group during the entire course of the experiment.
The following are detailed results of all treatment groups for selected time-points post-treatment, i.e. about 1 month and 1 year p.i. (also see Table 2).
On day 35 p.i., the highest amount of Gd was observed after treatment with the non-ionic linear GBCAs Omniscan (132 ± 23 nmolGd/g skin) and OptiMARK (47 ± 5 nmolGd/g skin). Lower Gd values in the skin were measured in animals treated with the ionic linear GBCAs Magnevist (36 ± 6 nmolGd/g skin) and MultiHance (7 ± 1 nmolGd/g skin). Even lower Gd values were observed in biopsies taken from animals injected with the macrocyclic GBCAs, ProHance (1 ± 1 nmolGd/g skin), Dotarem (2 ± 1 nmolGd/g skin) and Gadovist (2 ± 1 nmolGd/g skin) (Fig. 2a).
Comparison of gadolinium concentration in the skin biopsies taken from animals on day 35, 63, 168 and 364 p.i. after treatment with Omniscan, OptiMARK, Magnevist, Dotarem, ProHance and Gadovist. The animals were injected on five consecutive days with 2.5Â mmolGd/kg b.w. per injection and the Gd concentration in the skin is given as nmolGd/g skin
On day 364 p.i., the highest amount of Gd was, again, observed after treatment with the non-ionic linear GBCAs Omniscan (72 ± 12 nmolGd/g skin) and OptiMARK (18 ± 5 nmolGd/g skin). Lower Gd values in the skin were again measured in animals treated with the ionic linear GBCAs Magnevist (9 ± 2 nmolGd/g skin) and MultiHance (1.4 ± 0.4 nmolGd/g skin). For biopsies taken from animals treated with the macrocyclic GBCAs ProHance (0.08 ± 0.02 nmolGd/g skin), Dotarem (0.22 ± 0.17 nmolGd/g skin) and Gadovist (0.06 ± 0.03 nmolGd/g skin), the Gd values were in the same low range and close to the detection limit of the ICP-method as for untreated control animals (0.06 ± 0.03 nmolGd/g skin) and saline treated animals (0.18 ± 0.07 nmolGd/g skin) (Fig. 2d). For the Gd values measured 2 and 6 months p.i., please refer to Fig. 2b and c.
Statistical evaluation of the Gd concentration time-course in the skin tissue of rats of the treatment groups evaluated in set 1 (i.e. Omniscan, Magnevist, Gadovist, untreated) and set 3 (OptiMARK and saline) revealed clear differences between the various GBCAs. Figure 3 depicts the results of set 1 as box plots. No elevated Gd values could be detected in the skin tissue of rats treated with the macrocyclic compound, Gadovist, from day 24 onwards. However, after treatment with the three linear GBCAs Omniscan, OptiMARK and Magnevist elevated Gd levels in skin tissue were observed for the entire experimental observation time, with the non-ionic linear compound Omniscan, showing statistically significantly higher Gd levels compared with the ionic linear compound Magnevist.
Gadolinium concentration in the skin biopsies taken from animals treated with three different classes of GBCAs. The boxplots display the Gd content in the skin of treated animals after subtraction of the average Gd content of skin tissue of the respective control animals on the respective day and ten-times the corresponding standard deviation (SD). The null-line is set at the average value of Gd content of the skin of control animals plus ten-times the SD. The animals were injected on five consecutive days with Omniscan (upper panel), Magnevist (middle panel) and Gadovist (lower panel) at a dose of 2.5Â mmolGd/kg b.w. per injection and the Gd concentration in the skin is given as nmolGd/g skin
To estimate the Gd exposure to the skin, we calculated the AUC of the Gd values for the total observation period (until 364 days p.i., including the injection phase), as well as for the long-term retention period (‘plateau phase’) (63-364 days p.i.) (Fig. 4). Again, the highest total exposure was observed after treatment with the non-ionic linear GBCAs Omniscan (32.5 ± 4.4 molGd/g skin × 368 days and 24.4 ± 3.5 molGd/g skin × 301 days for the long-term retention period) and OptiMARK (9.4 ± 0.9 molGd/g skin × 368 days for the total Gd exposure of the skin and 7.2 ± 0.8 molGd/g skin × 301 days for the long-term retention period). Lower exposure of the skin to Gd was observed after treatment with the ionic linear GBCAs Magnevist (7.1 ± 0.4 molGd/g skin × 368 days for the total Gd exposure of the skin and 4.7 ± 0.3 molGd/g skin × 301 days for the long-term retention period) and MultiHance (2.2 ± 0.3 molGd/g skin × 368 days for the total Gd exposure of the skin and 1.5 ± 0.2 molGd/g skin × 301 days for the long-term retention period).
For all macrocyclic agents total Gd exposure in the skin (wash-out phase plus long-term retention phase) was higher than the saline and untreated controls. For the long-term retention period, however, exposure levels were in the same range as the controls (Fig. 4 and Table 2).
In general, the lowest values for the total exposure and for the exposure in the long-term retention period was observed after treatment with the macrocyclic GBCAs. Higher values were observed after treatment with the ionic linear GBCAs and the highest values were observed after treatment with the non-ionic linear GBCAs. Those differences were significant between treatment with macrocyclic GBCA Gadovist and the treatment with the non-ionic linear GBCAs Omniscan and OptiMARK. Additionally, those differences were also significant between treatment with the non-ionic linear GBCA Omniscan and treatment with the ionic linear GBCA Magnevist and macrocyclic GBCA Gadovist treatment. No statistical differences were found for the total exposure and the long-term retention period between any other two treatment groups.
Discussion
The aim of the study was to determine the elimination course and a potential long-term retention of Gd in the skin of rats after administration of different marketed GBCAs.
Regarding the elimination of Gd in the rat skin, in the immediate phase (wash-out phase), starting after the injection and lasting for several days (macrocyclic GBCAs) to weeks (linear GBCAs), the Gd concentration in the skin decreased considerably. This suggests that the Gd eliminated in this phase is readily eliminable, i.e. most likely water soluble and intact GBCA.
In the second phase (plateau phase), a steady-state level of Gd was observed, which decreased only minimally over time. For the linear GBCAs we observed Gd levels well above the level observed in control animals for up to a year after the last Gd injection (most pronounced in animals treated with Omniscan). For the macrocyclic GBCAs we observed the same low Gd levels in the skin as in control animals.
It has to be noted that we detected minor amounts of Gd in skin tissue of control animals, especially in set 1. This might be attributed to a potential Gd contamination in our laboratory. The level of contamination was in the same range as observed after administration of the macrocyclics, but well below the amount of Gd observed after administration of linear GBCAs (Table 2). Those differences in background values most likely also explain the minor differences in the Gd concentration and exposure of the skin to Gd observed after treatment with the different macrocyclic GBCAs.
The fact that the observed Gd was retained for such a long period in the skin of treated animals leads us to hypothesize that the Gd may exist in a water-insoluble state, which would make it less subject to elimination. But, as noted before, the methods used in this experiment to measure Gd levels could not distinguish between the various forms of Gd. Long-term retention of Gd in tissues has also been reported from NSF patients after multiple Gd-enhanced MRI investigations [22].
The time-course of the Gd concentrations in the skin (Fig. 1) and the exposure of the skin to Gd correlate with the stability of the GBCAs.
Based on kinetic considerations, the macrocyclic GBCAs are the most stable. In-vitro serum stability experiments could not detect any Gd release from macrocyclic agents within 15 days [23]. In the present experiment, all three macrocyclic GBCAs showed no long-term Gd retention in the skin tissue of treated rats that exceeded the level observed in control animals. All three macrocyclic agents showed virtually identical patterns of Gd elimination from the skin tissue of the study animals, reinforcing the notion that the minute in-vitro stability differences between the macrocyclic GBCAs are extremely unlikely to result in relevant differences in-vivo [3].
The ionic linear GBCAs are more stable than the non-ionic linear GBCAs [3], but both may release Gd in-vivo under certain circumstances. However, based on the results of in-vitro stability experiments in human serum [23] and buffered saline solution [24], any such release is expected to be significantly less with ionic linear agents than with non-ionic linear agents. Likewise, in our experiment, the Gd values observed on day 364 after the last injection of the ionic linear compound Magnevist were about nine-times lower compared with the treatment with the non-ionic linear compound Omniscan.
The non-ionic linear GBCAs have the highest level of Gd in the skin tissue of study animals in the plateau phase. Still, only a minute proportion of the injected Gd actually is retained in the body. The Gd values observed in the skin on day 364 p.i. of the Omniscan-treated animals correspond to about 0.081% of the total dose of Omniscan administered. This small fraction of Gd may explain why earlier pharmacokinetic studies failed to detect any Gd retention [25]. The difference in retained Gd between Omniscan and OptiMARK treatment may be related to the difference in the amount of excess ligand in the formulation of these two GBCAs [26].
While the difference in Gd retention is qualitatively correlated to the stability of the respective GBCA, no such correlation exists with the selectivity of the respective ligand to Gd over other metal ions or the acute toxicity of GBCAs, as indicated by LD50 values (Table 1). Such a hypothesis has been discussed previously [8, 9]. In contrast, our study suggests the importance of the stability of GBCAs to minimize Gd retention and to prevent the potential release of Gd.
This study has certain limitations. The experimental setting can be used to evaluate the concentration of Gd retained in rats after administration of different GBCAs using the dosing regimen employed in this study. However, due to significant differences in the pharmacokinetic properties of MultiHance in rats and humans, it is not possible to make valid, clinically relevant comparisons between MultiHance and other GBCAs using this model. Furthermore, as with all animal studies, caution must be exercised when extrapolating results to humans.
As the employed method for measuring Gd cannot distinguish between chelated and unchelated Gd, it is not possible to determine whether the retained Gd observed in the plateau phases after application of linear agents is chelated or unchelated. Furthermore, the study cannot answer whether the long-term retention of small amounts of the originally applied Gd doses plays any role in or contributes to the fibrosis observed in skin samples of NSF patients.
However, if Gd release plays a role in the onset of NSF, the data from this preclinical study suggest that there are potential differences in the risk profile for the different marketed GBCAs, with non-ionic linear agents possibly presenting a higher risk than the macrocyclic agents.
References
Caravan P, Ellison JJ, Mcmurry TJ, Lauffer RB (1999) Gadolinium(III) Chelates as MRI contrast agents: structure, dynamics, and applications. Chem Rev 99:2293–2352
Port M, Idée JM, Medina C, Robic C et al (2008) Efficiency, thermodynamic and kinetic stability of marketed gadolinium chelates and their possible clinical consequences: a critical review. Biometals 21(4):469–490, Aug
Schmitt-Willich H (2007) Stability of linear and macrocyclic gadolinium based contrast agents. Br J Radiol 80:581–582 author reply 584-585
Grobner T (2006) Gadolinium—a specific trigger for the development of nephrogenic fibrosing dermopathy and nephrogenic systemic fibrosis? Nephrol Dial Transplant 21:3604–3605
Marckmann P, Skov L, Rossen K et al (2006) Nephrogenic systemic fibrosis: suspected causative role of gadodiamide used for contrast-enhanced magnetic resonance imaging. J Am Soc Nephrol 17:2359–2362
Sieber MA, Pietsch H, Walter J, Haider W, Frenzel T, Weinmann H-J (2008) A preclinical study to investigate the development of Nephrogenic Systemic Fibrosis: a possible role for gadolinium-based contrast media. Inv Rad 43:65–75
Sieber MA, Lengsfeld P, Frenzel T, et al (2008) Preclinical investigation to compare different gadolinium-based contrast agents regarding their propensity to release gadolinium in vivo and to trigger nephrogenic systemic fibrosis-like lesions. Eur Radiol 18:2164–2173
Cacheris WP, Quay SC, Rocklage SM (1990) The relationship between thermodynamics and the toxicity of gadolinium complexes. Magn Reson Imaging 8:467–481
Mann JS (1993) Stability of gadolinium complexes in vitro and in vivo. J Comput Assist Tomogr 17(Suppl 1):S19–S23
Weinmann HJ, Press WR, Gries H (1990) Tolerance of extracellular contrast agents for magnetic resonance imaging. Invest Radiol 25(Suppl 1):S49–S50
Soltys RA (1992) Summary of preclinical safety evaluation of gadoteridol injection. Invest Radiol 27(Suppl 1):S7–S11
Wible JH Jr, Troup CM, Hynes MR et al (2001) Toxicological assessment of gadoversetamide injection (OptiMARK), a new contrast-enhancement agent for use in magnetic resonance imaging. Invest Radiol 36:401–412
Puttagunta NR, Gibby WA, Puttagunta VL (1996) Comparative transmetallation kinetics and thermodynamic stability of gadolinium-DTPA bis-glucosamide and other magnetic resonance imaging contrast media. Invest Radiol 31:619–624
Bianchi A, Calabi L, Giorgi C et al (2000) Thermodynamic and structural properties of Gd3+ complexes with functionalized macrocyclic ligands based upon 1,4,7,10-tetraazacyclododecane. J Chem Soc Dalton Trans:627–705
Clarke ET, Martell AE (1991) Stabilities of trivalent metal ion complexes of the tetraacetate derivatives of 12-, 13- and 14-membered tetraazamacrocycles. Inorg Chim Acta 37–49
Kumar K, Chang A, Francesconi LC et al (1994) Synthesis, stability, and structure of gadolinium(III) and yttrium(III) macrocyclic poly(amino carboxylates). Inorg Chem 33:3567–3575
Kumar K, Tweedle MF, Malley MF, Gougoutasi JZ (1995) Synthesis, stability, and crystal structure studies of some Ca2+ , Cu2+ , and Zn2+ complexes of macrocyclic polyamino carboxylates. Inorg Chem 34:6472–6480
Tóth É, Király R, Platzek J, Radüchel B, Brücher E (1996) Equilibrium and kinetic studies on complexes of 10-[2,3-dihydroxy-(1-hydroxymethyl)-propyl]-1,4,7,10-tetraazacyclododecane-1,4,7-triacetate. Inorganica Chimica Acta 249:191–199
Schmitt-Willich H, Brehm M, Ewers CL et al (1999) Synthesis and physicochemical characterization of a new gadolinium chelate: the liver-specific magnetic resonance imaging contrast agent Gd-EOB-DTPA. Inorg Chem 38:1134–1144
Skelly Frame EM, Uzgiris EE (1998) Gadolinium determination in tissue samples by inductively coupled plasma mass spectrometry and inductively coupled plasma atomic emission spectrometry in evaluation of the action of magnetic resonance imaging contrast agents. Analyst 123:675–679
Thomas R (2001) A beginner’s guide to ICP-MS. CRC Press, London
Thakral C, Abraham JL (2007) Automated scanning electron microscopy and x-ray microanalysis for in situ quantification of gadolinium deposits in skin. J Electron Microsc (Tokyo) 56:181–187
Frenzel T, Lengsfeld P, Schirmer H, Hütter J, Weinmann HJ (2008) Stability of gadolinium-based magnetic resonance imaging contrast agents in human serum at 37 degrees C. Invest Radiol 43:817–828
Laurent S, Elst LV, Muller RN (2006) Comparative study of the physicochemical properties of six clinical low molecular weight gadolinium contrast agents. Contrast Media Mol Imaging 1:128–137
Reinton V, Berg KJ, Savaland M, Andrew E, Normann PT, Rootwelt K (1994) Pharmacokinetics of gadodiamdie injection in patients with moderately impaired renal function. Acad Radiol 1:S56–S61
Sieber MA, Lengsfeld P, Walter J et al (2008) Gadolinium-based contrast agents and their potential role in the pathogenesis of nephrogenic systemic fibrosis: the role of excess ligand. J Magn Reson Imaging 27:955–962
Acknowledgements
This study was funded by Bayer Schering Pharma AG. The authors would like to thank Christiane Pering and Thomas Steger-Hartmann for the critical reading of the manuscript and Ines Heinzelmann, Stefanie Runge, and Andrea Baumgart for their excellent technical assistance. The authors would like to thank Carsten Schwenke, SCO: SSiS, Berlin, Germany, for his consultancy in statistics.
Open Access
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Open Access This is an open access article distributed under the terms of the Creative Commons Attribution Noncommercial License (https://creativecommons.org/licenses/by-nc/2.0), which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
About this article
Cite this article
Pietsch, H., Lengsfeld, P., Jost, G. et al. Long-term retention of gadolinium in the skin of rodents following the administration of gadolinium-based contrast agents. Eur Radiol 19, 1417–1424 (2009). https://doi.org/10.1007/s00330-008-1259-4
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00330-008-1259-4