Dynamic MRI for imaging tumor microvasculature: Comparison of susceptibility and relaxivity techniques in pelvic tumors

Authors


Abstract

Purpose

To assess the reproducibility of intrinsic relaxivity and both relaxivity- and susceptibility-based dynamic contrast enhanced (DCE) MRI in pelvic tumors; to correlate kinetic parameters obtained and to assess whether acute antivascular effects are seen in response to cisplatin- or taxane-based chemotherapy.

Materials and Methods

T1-weighted and T2*-weighted DCE-MRI and basal R2* measurements were performed on three consecutive days in women with gynecological tumors. The third scan was 21.0 (range 17.3–23.5) hours after the first cycle of chemotherapy. Kinetic parameter estimates were obtained and correlated between techniques. Test-retest reproducibility and response to treatment were assessed.

Results

Relative blood volume (rBV) and relative blood flow (rBF) correlated strongly with transfer constant (Ktrans), kep, and the initial area under the gadopentetate dimeglumine (Gd-DTPA) concentration-time curve (IAUGC) (all P < 0.01). The group 95% confidence interval (CI) for change was –10.8 to +12.1%; ±5.1%; –9.5 to +10.5%; ±7.5%; for Ktrans, ve, kep, and IAUGC, respectively, and ±13.6%, ±2.4%, ±11.6%, and ±11.0%, for rBV, mean transit time (MTT), rBF, and R2*, respectively. There were no significant acute changes in kinetic parameter estimates in response to treatment on group analysis, apart from a small decrease in ve.

Conclusion

The results confirm the dominant influence of flow on Ktrans in untreated gynecological tumors. There is no evidence of an acute, large magnitude antivascular effect caused by cisplatin- or taxane-based chemotherapy. J. Magn. Reson. Imaging 2007. © 2007 Wiley-Liss, Inc.

RELAXIVITY-BASED DYNAMIC contrast-enhanced (DCE) MRI (T1-weighted DCE-MRI) using low-molecular weight contrast media, has become established as a means of imaging tumor microcirculation in vivo (1) and, in particular, as a method of assessing response to therapy. Quantitative kinetic parameters that relate to tumor blood flow and permeability, such as the transfer constant, Ktrans, may be obtained by modeling contrast agent kinetics (2, 3) and both Ktrans and the initial area under the gadolinium concentration curve (IAUGC) have been recommended as biomarkers for assessing vascular response to treatment (4). Large magnitude antivascular effects of both vascular disrupting (angiolytic) and antiangiogenic agents in xenografts and in human tumors can be demonstrated using T1-weighted DCE-MRI (5–10).

A number of other MRI techniques can also be used to assess the functional tissue vasculature. Measurements of relative blood flow (and blood volume) can be obtained using bolus-tracking techniques with exogenously administered contrast media used as intravascular tracers with T2*-weighted DCE-MRI (susceptibility-enhanced MRI) (11). Measurements thus obtained have been found to correlate well with other imaging methods that can estimate tissue blood flow such as 133Xe–single photon emission computed tomography (SPECT) and 15O-H2O–positron emission tomography (PET). Brain tumor relative blood volume (rBV) measurements have been obtained using this technique (12–15) and correlate with measurements made with PET (15). More recently, T2*-weighted techniques have been applied to obtain estimates of relative blood flow (rBF) and rBV in visceral tumors (1, 16).

MRI using spoiled gradient-echo sequences with intrinsic susceptibility weighting (T2* or blood oxygenation level dependent [BOLD] weighting) reflect tissue flow and oxygenation status (17). T2*-weighted MRI is sensitive to changes in tumor oxygenation and blood flow in response to vascular stimuli. Changes in T2* relaxivity (R2* = 1/T2*) can be seen very early on following administration of a vascular disrupting agent and can thus be used as a biomarker of acute blood vessel shutdown (18, 19).

In this clinical trial, we tested the hypothesis that, unlike targeted vascular disrupting or antiangiogenic agents, no large antivascular effects are seen following conventional chemotherapy. T1-weighted, T2*-weighted DCE-MRI, and quantitative R2* techniques were used to investigate the acute antivascular effects of platinum and taxane-containing chemotherapy regimens. Patients were scanned on three consecutive days with chemotherapy given after the second scan, in order to determine test-retest variability of the technique (scans 1 and 2) and to assess acute response (scan 3). The clinical results for tumor T1-weighted DCE-MRI parameters Ktrans and IAUGC have been reported (20). In this report we expand on the reported T1-weighted DCE-MRI data by including leakage space (ve) and rate constant data (kep). T2*-weighted DCE-MRI and R2* data are also presented, with the T1-weighted DCE-MRI parameters included for comparison. We also report on the correlations between kinetic parameter estimates between T1- and T2*-weighted DCE-MRI kinetic parameters.

Thus, the purpose of this study was to: 1) document baseline values and assess reproducibility of intrinsic relaxivity and both relaxivity- and susceptibility-based DCE-MRI in pelvic tumors; 2) to cross correlate kinetic parameters obtained; and 3) to assess whether acute (<24-hour) antivascular effects are seen in response to cisplatin- or taxane-based chemotherapy.

MATERIALS AND METHODS

Patients and Treatments

Local ethics committee approval for the trial protocol and written informed consent from all participating patients were obtained. Eligibility criteria for the study were: histologically-confirmed pelvic cancer at an anatomical site suitable for imaging with MRI; tumor mass ≥3 cm in diameter; patient due to start first cycle of taxane- or platinum-based chemotherapy regimen; calculated creatinine clearance >50 mL·minute–1; World Health Organization (WHO) performance status ≤2; age ≥18 years; and no history of allergic reaction to contrast agents. Duration of taxane or platinum infusion varied from one hour (carboplatin, docetaxel) to six hours (cisplatin). Details of chemotherapy regimens used are given in Table 1.

Table 1. Patient Characteristics, Including Number of Imaging Episodes
Patient numberTumor typeTumor site imagedChemotherapy regimenaNumber of scans
  • *Chemotherapy given intravenously and repeated every three weeks unless otherwise stated.

  • a

    Carboplatin dose calculated according to AUC (area under the plasma-concentration time curve).

  • b

    T2-weighted sequence not performed posttreatment. Patients are ordered by mean pretreatment Ktrans.

  • S1 = pretreatment scan 1, S2 = pretreatment scan 2, S3 = posttreatment scan, PO = orally.

1Adenocarcinoma, ovaryPelvisCarboplatin AUC 6S1, S2, S3
2Mixed mullerian tumor/carcinosarcomaPelvisCisplatin 60 mg·minute–2; doxorubicin 60 mg·minute–2S1, S2, S3b
3Adenocarcinoma, ovaryPelvisCarboplatin AUC 6S1, S2, S3
4Adenocarcinoma, ovaryAnterior abdominal wallCarboplatin AUC 5; paclitaxel 175 mg·minute–2S1, S2, S3
5Adenocarcinoma, either from ovary or colonPara-aortic lymph nodesCarboplatin AUC 5S1, S2, S3
6Clear cell carcinoma, ovaryPelvisCarboplatin AUC 6S1, S2, S3
7Primary peritoneal carcinomaPelvisCisplatin 25 mg·minute–2; docetaxel 60mg·minute–2weeklyS1, S2, S3
8Mixed mullerian tumorPelvisCisplatin 70 mg·minute–2; epirubicin 70 mg·minute–2S1, S2, S3
9Adenocarcinoma, ovaryPelvisCisplatin 60 mg·minute–2; docetaxel 40 mg·minute–2weeklyS1, S2, S3b
10Adenocarcinoma, ovaryPelvisDocetaxel 80 mg·minute–2day 1; gemcitabine 1250 mg·minute–2days 1 and 8S1, S2, S3
11Granulosa cell tumor, ovaryPelvisBleomycin 30 mg days 2,8,15; etoposide 165 mg·minute–2days 1-3, Cisplatin 50 mg·minute–2 days 1 and 2S1, S2, S3
12Adenocarcinoma, ovaryPelvisCarboplatin AUC 6S1, S2, S3
13Adenocarcinoma, ovaryPelvisCarboplatin AUC 6S1, S2, S3
14Primitive neuroectodermal tumorPelvisCisplatin 50 mg·minute–2 days 1 and 2; etoposide 150 mg·minute–2 days 1, 2, and 3S1, S2, S3
15Adenocarcinoma, ovaryPelvisCarboplatin AUC 6S1, S2, S3
16Adenocarcinoma ovaryPelvisCisplatin 60 mg·minute–2weekly; etoposide 50 mg PO for every 21/28 daysS1, S2, S3
17Primary peritoneal carcinomaPelvisCarboplatin AUC 6; paclitaxel 175 mg·minute–2S1, S2, S3
18Adenocarcinoma, endometriumPelvisCisplatin 60 mg·minute–2; doxorubicin 60 mg·minute–2S1, S2, S3
19Squamous cell carcinoma, cervixCervixCisplatin 60 mg·minute–2; bleomycin 30 mg·minute–2; methotrexate 300 mg·minute–2S1, S2
20Poorly differentiated carcinoma ?ovary ?primary peritoneal carcinomaAnterior abdominal wallCarboplatin AUC 5S1, S2

MRI Data Acquisition

Three DCE-MRI scans were performed on consecutive days: two prechemotherapy to assess the reproducibility of the technique and one 20 to 24 hours after the start of the first cycle of chemotherapy to assess response. MRI studies were performed on a 1.5-T Magnetom Symphony scanner (Siemens Medical Systems, Erlangen, Germany), using a body phased-array coil. In the first scanning session, initial whole pelvis T1- and T2-weighted anatomical images were obtained to select four suitable contiguous slices through the center of a tumor mass. Care was taken to place the patient in a similar bed position on the follow-up sessions, to reproduce as far as possible the original anatomical slice location. Repeat scans were always performed by the same radiographer with a physicist and research fellow in attendance in order to ensure repositioning reproducibility. The following MR sequences were then acquired in order:

  • 1Multiple spoiled gradient-recalled echo (GRE) images at a single tumor slice level obtained with increasing TE times (TE = 5–75 msec, TR = 100 msec, flip angle (α) = 40°, slice thickness = 8 mm, field of view [FOV] = 350 mm, reconstruction matrix size = 2562, in-plane resolution = 1.37 × 2.19 mm).
  • 2An intermediate weighted (proton density [PD]-weighted) spoiled GRE sequence: TE = 4.7 msec, TR = 350 msec, α = 6°, slice thickness = 8 mm, FOV = 350 mm, reconstruction matrix size = 2562, in-plane resolution = 1.37 × 2.19 mm; image slices = 4 (one of which coincided with the slice position in 1 above).
  • 3A dynamic series of 160 T1-weighted GRE images: TE 4.7 msec, TR 11 msec, α = 35°, otherwise as in 2 above; one set of four images (registered to the PD image positions) acquired every 12 seconds for a total imaging time of eight minutes and five seconds. Gadopentetate dimeglumine (gadolinium diethylenetriamine penta-acetic acid [Gd-DTPA], Magnevist®; Schering Health Care Ltd., Burgess Hill, UK) was the contrast agent used, injected intravenously using a power injector (dose 0.1 mmol·kg–1 of body weight) at 4 mL·second–1 during the fifth acquisition. System gain and scaling factors were maintained between acquisition of the PD- and T1-weighted dynamic series of images, to enable the calculation of tissue T1-relaxivity and contrast agent concentration (21).
  • 4A dynamic series of 60 T2*-weighted spoiled GRE images was then acquired from a single slice (TE = 20 msec, TR = 30 msec, α = 40°, slice thickness = 8 mm, FOV = 350 mm, in-plane resolution = 2.73 × 2.73 mm, time resolution = 1.9 seconds), registered to the image in 1 above. A second Gd-DTPA injection (dose = 0.2 mmol·kg–1, rate = 4 mL second–1) was administered after the 10th acquisition.

Image Analysis

Images were transferred to a Sun Ultra 60 workstation, (Sun Microsystems, Mountain View, CA, USA) and analyzed using specialist software (Magnetic Resonance Imaging Workbench [MRIW], Institute of Cancer Research, London, UK) (22). Using information from anatomical and postcontrast T1-weighted DCE-MRI subtraction images, regions of interest (ROIs) were carefully drawn around the tumor edges on all slices by a radiologist with eight years of experience in DCE-MRI, who carefully excluded areas of artifacts and blood vessels. Similar ROIs were used for all three MR examinations in an individual patient, all of which were drawn at the same session. The consistency of ROI placement for this expert reviewer has been documented at a 3.7% variation in terms of ROI size definition (23).

Intrinsic Susceptibility-Weighted Analysis

R2* maps were calculated using in-house software written in IDL® (Research Systems, Inc., Boulder, CO, USA). A straight line was fitted to a plot of lnSt against TE for each voxel using a least-squares approach, of which the gradient is –R2* (units second–1). Voxels with either negative or zero values were excluded from analysis.

T1-Weighted DCE-MRI Analysis

MRIW software was used to convert MRI signal intensities of the T1-weighted DCE-MRI dataset into T1 relaxation rates (R1) and then into Gd-DTPA concentrations for individual voxels, using the methods described by Parker et al (21). The following T1-weighted parameters were calculated in the MRIW software: the initial area under the Gd-DTPA concentration-time curve for the first 60 seconds (IAUGC; mM·second), transfer constant (Ktrans; minute–1), rate constant, (kep, minute–1), and leakage space (ve, %). The changing tissue Gd-DTPA curve was fitted to a standard compartmental model (24), to characterize the influx of Gd-DTPA into the tumor extracellular extravascular space and its venous efflux, to obtain Ktrans, kep, and ve. A pooled arterial input function (AIF) was used for the modeling procedure (3, 25) as described previously (26).

Voxels that did not enhance and those enhancing voxels that failed the modeling process or had values >5.0 minute–1 were excluded from statistical analysis for the Ktrans, ve, and kep parameters. However, IAUGC estimates include voxels that failed the modeling process as these will include Ktrans >5 minute–1. Statistical analysis was performed on combined voxel data from all slices, taking the median voxel value as representative of central tendency. Median rather than the mean voxel values were used, as histogram distributions of some kinetic parameters were skewed.

T2*-Weighted DCE-MRI Analysis

Changes in T2 relaxivity (ΔR2*) were calculated for each time point of the T2*-weighted DCE-MRI dataset and fitted using a gamma (Γ)-variate function (27, 28) using MRIW software:

equation image(1)

where C(t) is contrast agent concentration in blood at time t, S0 and S(t) are the signal intensities at the baseline and at time t, respectively, and TE is the echo time of the MR sequence used (11). The rBV (arbitrary units) is then the integral of the ΔR2*-time curve (using the Γ-variate fit).

equation image(2)

The relative mean transit time (MTT, second) was approximated by measuring the width of the ΔR2*-time curve at half its maximum value (full-width, half-maximum) (29). rBF (arbitrary units) was then obtained by substituting in the transit time equation of the central volume theorem

equation image(3)

A single global value for the entire ROI was obtained by taking the median of all the individual voxel values of the parameters (the median was used rather than the mean because of skewed distributions).

Statistical Analysis

We used the standard consensus approach to assessing reproducibility as described by Bland and Altman (30, 31). The 95% confidence interval (CI) is the key statistical parameter that is used to determine whether a change in a kinetic parameter following an intervention is statistically significant or not for one or several patients (31).

For each patient, the difference d between the two pretreatment measurements of a parameter was calculated. Data were transformed using natural logarithms if the variability of d was found to depend on its mean value (a significant two-tailed Kendall's tau test) (30). The square root of the mean squared difference, dsd, (=√[(Σd2)/n] where n is the number of patients) was then calculated. The 95% CI for change for a group of n patients is then equal to ±(1.96 × dsd)/√n). For an individual patient, n = 1 so the 95% confidence interval for change is equal to ±(1.96 × dsd), which is also known as the repeatability statistic, r (31).

The within-patient SD, wSD = dsd/√2, is calculated and used to estimate the within-patient coefficient of variation, wCV, by dividing wSD by the group mean pretreatment value for each parameter. wCV quantifies measurement error relative to the size of the (positive) kinetic parameters. If data had to be transformed, then wCV was approximated by wCV = ewSD – 1 (30).

The results of the reproducibility analysis were then used to assess whether there had been a statistically significant change in kinetic parameters due to chemotherapy, either for individual patients or the group. As there were two pretreatment measurements (days 1 and 2), the mean of them was taken as the pretreatment value for each parameter. Our hypothesis was that there would be no difference between this combined pretreatment value and the posttreatment value. For individual patients, the repeatability statistic, r, expressed as a percentage of the group mean pretreatment value for each parameter, gives a range within which the difference between pre- and posttreatment values would be expected to lie for 95% of observations, assuming that the hypothesis is true. If the difference falls outside this range for a particular kinetic parameter, then a significant change was deemed to have occurred. Similarly, to assess mean response in the group, the 95% CI for change, expressed as a percentage of the group mean pretreatment value, gives the range required.

In addition, Spearman's ρ (rho) was used to determine whether there was a significant correlation between any of the baseline (day 1) kinetic parameter values (JMP Statistics, version 3.2.6; SAS Institute, Inc., Cary, NC, USA).

RESULTS

A total of 24 women with gynecological tumors were imaged. Data from four patients were excluded from analysis (three technical failures, one voluntary patient motion) leaving 20 women for data analysis (mean age = 57 years; range = 29–74 years). For the T1-weighted DCE-MRI study there were 20 complete data sets for the reproducibility analysis and 18 complete data sets for the response analysis (data could not be included from one patient due to internal organ motion and one patient was unable to complete the post-treatment scan, due to treatment toxicity). For the T2*-weighted DCE-MRI study there were 16 complete data sets for reproducibility and response analysis, as four patients did not complete this section on at least one of the imaging days (one due to concern that the intravenous cannula had tissued and three due to claustrophobia). There were 20 complete data sets for the R2* reproducibility and response assessments.

Table 1 shows patient characteristics including treatments received and the number of scans for each patient. The baseline (day 1) means of the median kinetic parameter values for tumor ROI were as follows: Ktrans = 0.393 minute–1 (range = 0.192–1.081); ve = 40.8% (range = 27.1–65.2); kep = 0.967 minute–1 (range = 0.449–2.684); IAUGC = 6.39 mM·second (range = 2.14–12.1); rBV = 263.7 arbitrary units (AU) (range = 59.8-483), MTT = 24.7 second (range = 21.2–28.6); rBF = 10.5 AU (range = 1.46–19.15); and R2* = 20.8 seconds–1 (range = 9.1–36.5). Table 2 shows the reproducibility analysis for all measured parameters for tumor for T1-weighted and T2*-weighted parameters.

Table 2. Reproducibility Analysis for Tumor T1- and T2*-Weighted Parameters
 T1-weighted parametersT2*-weighted parameters
Ktrans (minute–1)ve (%)kep (minute–1)IAUGC (mM·second)rBV (AU)MTT (second)rBF (AU)R2* (second–1)
  • T1-weighted parameters: four-slice concatenated data, T2*-weighted parameters: single-slice data. Ktrans and kep data are logarithmically transformed.

  • Mean = group mean pretreatment value, dsd = square root of the mean squared difference, r = individual patient repeatability, r (%) = repeatability as a % of group mean pretreatment value, CI = 95% confidence interval for change as a % of group mean pretreatment value, wCV = within-patient coefficient of variation.

Mean0.3940.60.956.36233.424.99.5321.6
dsd0.264.740.231.0964.981.192.265.44
r0.519.290.442.14127.32.344.4210.6
r (%)–40.0 to +66.7%± 22.9%–35.9 to +55.9%± 33.7%± 54.5%± 9.4%± 46.4%± 48.6%
CI (%)–10.8 to +12.1%± 5.1%–9.5 to +10.5%± 7.5%± 13.6%± 2.4%± 11.6%± 11.0%
wCV (%)20.3%8.3%17.4%12.1%19.7%3.4%16.8%17.5%

Table 3 shows the correlation between different parameters. IAUGC was found to correlate very strongly with Ktrans (r = 0.86; P < 0.0001) (Fig. 1) and strongly with ve (r = 0.68; P = 0.01). rBV and rBF correlated strongly with each other (r = 0.96; P < 0.0001) reflecting the narrow range of MTT values observed (see above). There were also strong correlations between T1-weighted (Ktrans, kep, and IAUGC) and T2*-weighted DCE-MRI parameters values (especially rBF). These correlations between tumors were also observed spatially within tumors on corresponding voxel maps (Fig. 2).

Table 3. Correlation Coefficients for T1- and T2*-Weighted DCE-MRI and R2* Parameters
 KtransvekepIAUGCrBVMTTrBF
  • Correlations within T1-weighted and T2*-weighted DCE-MRI parameters are not given, as they are related mathematically (Ktrans/ve = kep and rBF = rBV/MTT).

  • NS = not significant.

IAUGC0.86; P < 0.00010.68; P = 0.010.58; P = 0.01    
rBV0.69; P < 0.0010.28; NS0.59; P = 0.010.55; P = 0.01   
MTT–0.40; NS–0.14; NS–0.34; NS–0.40; NS   
rBF0.77; P < 0.00010.33; NS0.64; P < 0.010.65; P < 0.01   
R2*0.51; P < 0.050.25; NS0.54; P = 0.010.44; NS0.23; NS–0.07; NS0.28; NS
Figure 1.

Correlation of transfer constant (Ktrans) with IAUGC and relative blood flow (rBF). Scatter plots and simple linear regressions of baseline (day 1) values of transfer constant (Ktrans) with IAUGC and relative blood flow in 19 patients showing close correlations. The transfer constant outlier (patient 20) shown as a solid square block for IAUGC and as a crossed square block for rBF was not included in the regression plot.

Figure 2.

Recurrent ovarian cancer. a:Patient #9 with recurrent ovarian cancer. (A) T2-weighted and (B) 100-second subtraction images are shown for reference. (C) Ktrans map (color scale = 0–1 minute–1), and (D) rBF map (color scale = 0–40 AU). Note that visually there is a close spatial correlation between Ktrans and rBF. The sites of five ROIs are marked on the subtractions image (B) and corresponding kinetic curves are shown in part b. b: Each graph shows T1- and T2*-weighted DCE-MRI curves for the five ROIs indicated in part a (B) superimposed on the same time scale. Corresponding calculated values of Ktrans and rBF for these ROIs are also given. The zero point on the time scale represents the point of injection of contrast medium for both studies which were performed consecutively. The earlier onset and shorter duration of T2*-weighted DCE-MRI which occurs just before the upslope on the T1-weighted enhancement curves for all ROIs except ROI-2 confirms that the upslope of T1-weighted DCE-MRI has a significant vascular contribution. ROI-2 represents a cystic area seen clearly in part a (A); here the flow contribution is undetectable by both T1- and T2*-weighted DCE-MRI. The simple linear regression plot is of Ktrans vs. rBF for the same ROIs (error bars represent 1 SD) confirms the strong spatial correlation of Ktrans with rBF.

Figures 3 and 4 show tumor pretreatment values (mean of the two pretreatment examinations), the posttreatment values and repeatability ranges for all parameters obtained for each patient. There were no significant changes posttreatment in Ktrans for any patient. One patient had a significant decrease in ve (#11) and one patient had a significant increase in kep (#7). Two patients had significant increases (#2 and #10) but one patient had a significant decrease (#11) in IAUGC. Three patients had a significant increase in rBV (#7, #15, and #17) and one had a significant decrease in rBV (#1). The same three patients had an increase in rBF (#7, #15, and #17) and two had a significant decrease (#1 and #16). One patient had a significant increase (#16) and one patient a significant decrease (#1) in MTT. One patient had a significant increase in R2* (#1).

Figure 3.

Individual patient data for T1-weighted parameters. a: Ktrans; b: ve; c: kep; d: IAUGC. The mean pretreatment values (gray circle), posttreatment value (black square), and the repeatability range for each parameter are shown. Significant day 3 values are marked with an arrow. Patients are ordered by mean pretreatment Ktrans.

Figure 4.

Individual patient data for T2-weighted parameters. a: rBV; b: MTT; c: rBF; d: R2*. The mean pretreatment values (gray circle), posttreatment value (black squares) and the repeatability range for each parameter are shown. Significant day 3 values are marked with an arrow. Patients are ordered by mean pretreatment Ktrans. # = missing data.

When patients were analyzed as a group, there were no significant changes in Ktrans, kep, or IAUGC (mean % change = +0.32%, +7.8%, –3.0%, respectively; see Table 2 for group 95% CIs) but there was a 5.5% decrease in ve that was significant. There were no significant changes in rBV, MTT, rBF, or R2* (mean % change = +8.5%, –1.0%, +8.6%, and +3.5%, respectively; see Table 2).

DISCUSSION

As far as we are aware there are no literature data documenting the values of kinetic parameters using T2*-weighted DCE-MRI or R2* in visceral tumors in humans nor the level of test-retest variability for these parameters. Furthermore, the cross-correlation between kinetic parameters obtained by DCE-MRI techniques that utilize different contrast mechanisms (susceptibility- and relaxivity-based techniques) or with intrinsic relaxivity values in pelvic tumors has not been reported previously. We have also shown for the first time that are no large magnitude antivascular effects detected within 24 hours of the first dose of chemotherapy in any of these metrics in group analysis. Each of these aspects of the study will be addressed in turn.

The values of the T1-weighted DCE-MRI parameters are in a similar range to those found previously in visceral tumors (26). The T2*-weighted parameters (rBV, rBF, and MTT) are relative because the absolute input function is not known and so cannot be reliably compared between studies. However, baseline R2* values may be compared to values quoted in the literature for 1.5-T systems. The mean baseline (day 1) R2* value was 20.8 (±1.6) second–1. Rijpkema et al (32) reported a mean baseline R2* of 29.56 (±1.29) second–1 for 11 patients with head and neck squamous cell carcinomas (also measured using a 1.5-T system). The differences between R2* values may be due to the higher levels of hypoxia seen in head and neck squamous cell carcinomas which would lead to increased R2* (17).

We have calculated reproducibility for the T2*-weighted flow parameters. Jackson et al (33, 34) have performed a reproducibility analysis for rBV for 11 patients with glioma (scanned 36–56 hours apart) and for five patients with hepatic neoplasms (scanned 48–56 hours apart). Unfortunately, their analysis is not directly comparable with our data as different reproducibility statistics are quoted.

IAUGC was found to correlate very strongly with Ktrans and strongly with ve—reflecting its value as a “summary” T1-parameter. IAUGC was also found to closely correlate with rBF/rBV. This shows that the IAUGC parameter, which can be obtained without mathematical modeling of the plasma Gd-DTPA concentration, is a useful kinetic parameter; reinforcing the recommendation for its usage for evaluating antivascular anticancer therapies (4). However, it is important to remember that parameters such as transfer constant provide physiological information about contrast agent behavior in tissues, whereas the relationship of IAUGC to vascular physiology is complex and not specifically related to vascularity as it is also affected by flow and extracellular space (35). The close correlation between rBV and rBF is expected because the parameters are related mathematically and is as expected because of the narrow range of MTT. The narrow range of MTT is intriguing (and quite different from what has been found in stroke patients, in which MTT is typically prolonged in regions of brain ischemia) suggesting that arteriovenous shunting at the microscopic level may be considerable in these tumors.

In the assessment of tumor response, there were no significant changes observed in group analyses for T2*-weighted parameters but a few individuals had either significant increases or decreases in certain parameters. It is possible that the changes seen are due to chance rather than real events. A statistical correction factor can be applied for multiple-parameter comparisons—the Bonferroni correction (36). However, application of the Bonferroni correction would be too conservative in this instance, as the measured parameters are not independent of each other.

A demonstrable antivascular effect should result in reductions of the parameters. Changes in transfer constant and IAUGC have been reported by us elsewhere, and are included in this report for comparison (20); in this report we expand on the reported T1-weighted DCE-MRI data by including leakage space and rate constant data. In individual patient analyses there were no significant decreases in transfer constant and only one patient had a significant decrease in IAUGC (#16; decrease of 39.3%). There were no significant changes in group Ktrans or IAUGC. It has been shown previously that significant antivascular effects cause reduction in these parameters (6, 8). The reason for the small reduction in group leakage space is uncertain. In the one individual who had a significant reduction in leakage space, the magnitude of reduction was only just outside the repeatability range.

It has been shown previously that R2* can be used as a biomarker for monitoring the acute (within hours) effects of vascular disrupting agents (18, 19). Robinson et al (19) detected an increase in R2* (consistent with an increase in hypoxia) at 35 minutes after administration of the vascular targeting agent, ZD6126, in an animal model. Gross et al (18) investigated the effects of a novel photodynamic therapy drug (that acts as a vascular disrupting agent) in melanoma xenografts using BOLD-MRI. An increase in R2* was seen within a few minutes of illumination (to activate the drug)—consistent with oxygen consumption by the photochemical reaction and acute blood vessel shutdown (18). However, recent work also shows that R2* measurements made at 24 hours following an antivascular intervention are probably not likely to be very helpful, as the presence of necrosis and/or edema will influence R2* measurement (7, 37). We were unable to show group changes in R2* posttreatment with chemotherapy but one patient had a significant increase.

We have used a number of MRI methods to try to detect the antivascular effects of platinum/taxane chemotherapy with 24 hours of the first dose of treatment and we have found no systematic effects. However, it is not possible to state categorically that the cytotoxic agents tested have no acute effects on tumor vascularity (if they are present, then they are of small magnitude and beyond the resolution of the techniques that we have used). It may be that compared with antiangiogenic or vascular disrupting agents, conventional cytotoxins take longer to have an effect, as often they will only take effect as the cell enters the cell cycle, and this will therefore depend on endothelial cell turnover or cycle time. The use of T1-weighted DCE-MRI to detect antivascular effects of agents targeted against angiogenesis or the vascular endothelium has been recommended by international consensus panel meetings (4) and in reports from the National Cancer Institute (http://imaging.cancer.gov/reportsandpublications/ReportsandPresentations/MagneticResonance). These panels have recommended further research on the usage of other imaging techniques and the data presented here adds to this knowledge base. The usage of T2* DCE-MRI to obtain measurements of relative cerebral blood flow and blood volume has been validated but there is uncertainty regarding its application for the measurement of rBF and rBV in visceral tumors. This is because Gd-DTPA leaks rapidly out of tumor vessels during the first pass of contrast media, resulting in T1 “shine through” effects. These effects can be partially overcome by using a gamma-variate function, fitted to the changing R2* data and by preenhancing the tumor with a dose of contrast agent prior to the T2* DCE-MRI data acquisition (14), which we have done. We recognize that the methods that we have used are likely to lead to an overestimation of the true MTT and an underestimation of rBF (29).

There are also uncertainties with regard to the reliability of kinetic parameter estimates derived from the application of tracer kinetic models to T1-weighted DCE-MRI data. These derive from assumptions implicit in kinetic models and those for the measurement of tissue contrast agent concentration. For example, the Tofts' model uses a standard description of the time varying blood concentration of contrast agent, and assumes that the supply of contrast medium is not flow limited and that tissue blood volume contributes negligibly to signal intensity changes compared with that arising from contrast medium in the interstitial space. Reliable methods for measuring arterial input function for routine DCE-MRI studies are only now emerging and were not widely available at the time when this study was conducted. Buckley (38) has suggested that the application of commonly accepted models and their respective model-based assumptions to DCE-MRI data leads to systematic overestimation of Ktrans in tumors. It is difficult to be certain about how accurately model-based kinetic parameter estimates compare with the physiological parameter that they purport to measure, particularly as there is no reliable clinical gold standard. Nevertheless, there is a growing literature base that supports the notion that Ktrans is flow-dominated in untreated tumors (3, 39, 40) and the results presented here add to this literature.

In conclusion, we have performed a mechanistic clinical trial to understand the nature of tumor vascular response to conventional chemotherapy. In doing so, we have established the intrinsic variability of kinetic parameters estimated derived from susceptibility- and relaxivity-based dynamic MRI techniques in human gynecological tumors. We have independently confirmed that transfer constant is flow-dominated in untreated female pelvic tumors. We have also shown that no large magnitude, acute antivascular effects are seen within 24 hours after the first dose of platinum/taxane chemotherapy. It is important to remember that it is not possible to categorically state that the cytotoxic agents tested have no acute effects on tumor vascularity (if they are present, then they are of small magnitude and beyond the resolution of our techniques). Therefore, if significant acute (<24 hours) reductions in kinetic MRI parameters are seen in combination cytotoxic and antivascular therapies, one may presume they are due to the vascular-directed agent. This is also the first demonstration that cytotoxic agents differ from vascular-disrupting agents in their acute effect on DCE-MRI parameters in humans, thus validating the use of DCE-MRI as a biomarker of vascular targeting activity.

Acknowledgements

We thank Professor Søren Bentzen and Dr. Elena Kulinstaya for statistical advice.

Ancillary