Thermal dose fractionation affects tumour physiological response

Abstract

Purpose: It is unknown whether a thermal dose should be administered using a few large fractions with higher temperatures or a larger number of fractions with lower temperatures. To evaluate this we assessed the effect of administering the same total thermal dose, approximately 30 CEM43T₉₀, in one versus three to four fractions per week, over 5 weeks.

Materials and methods: Canine sarcomas were randomised to receive one of the hyperthermia fractionation schemes along with fractionated radiotherapy. Tumour response was based on changes in tumour volume, oxygenation, water diffusion quantified using MRI, and a panel of histological and immunohistochemical end points.

Results: There was a greater reduction in tumour volume and water diffusion at the end of therapy in tumours receiving one hyperthermia fraction per week. There was a weak but significant association between improved tumour oxygenation 24 h after the first hyperthermia treatment and extent of volume reduction at the end of therapy. Finally, the direction of change of HIF-1α and CA-IX immunoreactivity after the first hyperthermia fraction was similar and there was an inverse relationship between temperature and the direction of change of CA-IX. There were no significant changes in interstitial fluid pressure, VEGF, vWF, apoptosis or necrosis as a function of treatment group or temperature.

Conclusions: We did not identify an advantage to a three to four per week hyperthermia prescription, and response data pointed to a one per week prescription being superior.

• thermal dosimetry
• tumor oxygenation
• microenvironment

Introduction

After decades of investigation into the biology and physical principles of hyperthermia, randomised clinical trials have shown that hyperthermia led to improved local control of the following irradiated human tumours compared to radiation alone: melanoma [1], glioblastoma multiforme [2], breast cancer [3], head and neck tumours [4], pelvic tumours [5] and superficial tumours [6]. Despite these successes, the application of radiation and hyperthermia for treatment of solid tumours has not been adopted widely. One reason for this is the lack of well-defined thermal dosimetry goals. Even today, many clinical hyperthermia treatments are undertaken without a priori establishment of a hyperthermia dose prescription. Hyperthermia may be the only cancer treatment modality where that philosophy is accepted. We know that the total cumulative thermal dose is related to duration of tumour control [6, [7]] and it is also very likely that there are differences in tumour response depending on how the total cumulative thermal dose is fractionated. Until the fractionation effect is understood and thermal dose is defined and administered prospectively, clinical trials will be suboptimal and the cancer community's confidence in hyperthermia treatments will remain tentative.

Current day thermal dose prescriptions may be ill-defined because of the need for invasive thermometry to characterise the temperature distribution adequately. Wider application of non-invasive magnetic resonance (MR) thermometry would solve this problem. However, establishing a non-invasive MR thermometry programme is time, labour and cost intensive. In the meantime, many hyperthermia treatments will continue to be assessed using invasive thermometry. This does not mean that thermal prescriptions should not be defined, as administering a prescribed thermal dose accurately based on limited thermometry is possible [7, [8]], and, as already noted, thermal dose–response relationships based on invasive thermometry have been characterised in canine [7] and human [6] tumours. Nevertheless, thermal dose relationships will be even more evident when non-invasive thermometry is more widespread.

Even though thermal dose can be delivered as prescribed, the optimal time–dose prescription is not known. Both cytotoxicity and alteration of tumour oxygenation, likely the major mechanisms by which hyperthermia exerts its anti-tumour effect, are a function of the time–temperature relationship. For example, higher temperatures are more likely to lead to cytotoxicity [9, [10]] but these could also have a deleterious effect on tumour vasculature [11], leading to hypoxia, radio-resistance and phenotypic aggressiveness. On the other hand, lower temperatures would be less cytotoxic but might result in improved perfusion and oxygenation [11–13] with better drug delivery and a reduction in tumour hypoxia that leads to improved radiation response. It remains to be determined which of these strategies is optimal in the clinic. For example, should higher temperatures be achieved over a few fractions with the aim being to kill tumour cells, or should the temperature be reduced intentionally and a larger number of fractions administered to take advantage of improved tumour oxygenation? This is a critical question and the answer could have a profound impact on how hyperthermia should be administered in human treatments. Our goal was to assess this by characterising the physiological effect of two different fractionation schemes in spontaneous canine sarcomas where the same total cumulative thermal dose was administered.

Materials and methods

This study was conducted in spontaneous canine soft tissue sarcomas, which include undifferentiated sarcoma, fibrosarcoma, myxosarcoma, haemangiopericytoma, liposarcoma and neurofibrosarcoma. Extremity tumours are most common but canine soft tissue sarcomas also arise on the head, oral cavity, and trunk. They are locally invasive into soft tissue, but rarely invade bone, except in the oral cavity. Canine soft tissue sarcomas have metastatic potential but metastasis does not occur with such a high frequency or sufficiently early to interfere with evaluation of the primary tumour. Tumours were measureable and had a volume between 10 cm³ and 400 cm³ and there was no evidence of local bone invasion in any dog.

Before treatment, dogs underwent quantification of tumour volume (using calipers), tumour oxygenation and interstitial fluid pressure, followed by tumour biopsy. Pretreatment diffusion-weighted MRI was performed in some dogs after the project had begun.

Multiple pretreatment tumour biopsies were acquired with a Tru-Cut biopsy needle (UK Medical, Sheffield) and archived. The Oxford Optronix Oxylite system was used for oxygen measurement (http://www.oxford-optronix.com/ptissuemonitoring.htm).

The flexible probes, 250–500 µ in diameter, were inserted through a pre-placed catheter and measurements were recorded in 5-mm increments, or less in small tumours, as the probe was withdrawn. A minimum of 20 oxygen determinations was sampled from four tracks, as recommended by others [14]. Tumour oxygenation was expressed as mean pO₂, median pO₂, % measured points ≤2.5 mm Hg, % measured points ≤5 mm Hg and % measured points ≤10 mm Hg.

Interstitial fluid pressure was measured with a wick and needle system [15]; four sites were measured at each time point. Interstitial fluid pressure was expressed as mean and median values.

Water diffusion was quantified using diffusion weighted imaging (DWI) with a 1.5 T magnet (Siemens Symphony, Siemens Medical Systems, Malvern, PA). Dogs were under general anaesthesia, breathing isoflurane in 100% oxygen, for MR imaging. DWI was performed using a half Fourier acquisition, diffusion-weighted, single shot turbo spin echo (HASTE) sequence with b values of 0 and 500 s/mm² (TR 3,000 ms, TE 132 ms, echo train length 256, 128 × 128 matrix, number of excitations (NEX) = 2 for b = 0 s/mm² and NEX = 6 for b = 500 s/mm², 4.0 mm slice thickness, 0.5 mm gap). Diffusion-weighted images were used to compute the apparent diffusion coefficient (ADC) using the magnet's proprietary software. T2-weighted spin echo images were also acquired and used for anatomic registration of the region of interest (ROI) used for quantification of the ADC. The ADC was quantified in every pixel in the tumour by drawing a ROI around the tumour in every T2-weighted image and then transferring the ROI to the corresponding ADC image. ADC was quantified using ImageJ [16] (http://rsbweb.nih.gov/ij/).

After pretreatment measurements, dogs received local hyperthermia on a subsequent day, under general anaesthesia, using a stationary or scanning spiral 433 MHz superficial microwave applicator coupled to the skin with deionised water. We have used these applicators to heat canine tumours for many years. The maximum tumour volume in this study was 400 cm³, according to protocol design. We have heated tumours of this volume satisfactorily in prior studies with the equipment used herein. Additional details on the hyperthermia technique can be found elsewhere [7]. Thermometry probes were translated automatically through pre-placed intra-tumoural catheters by computer control to record temperatures at 0.5–1.0 cm increments across the tumour at 3–5 min intervals. Generally, temperature was monitored at approximately 20 discrete points from two to four catheters. Thermal dose descriptors were calculated according to standard thermal isoeffect dose relationships [10, [17]] using software designed at Duke University. The target cumulative thermal dose was between 20–50 CEM43T₉₀, with the aim of being as close to 50 CEM43T₉₀ as possible at the end of the treatment course. The dose of 20–50 CEM43T₉₀ was the target because it was the dose associated with significant improvement in local tumour control compared to a dose of 2–5 CEM43T₉₀ in a prior canine soft tissue sarcoma study [7].

Heating a tumour with an external power source results in temperature heterogeneity that fluctuates with time, between treatments and between tumours. This heterogeneity creates problems in quantifying thermal dose for accurate reproduction of dose from treatment to treatment. When combined with radiation, the beneficial effects of hyperthermia appear related to the lower temperatures in the heterogeneous temperature distribution [18, [19]]. To quantify thermal dose from a heterogeneous temperature distribution, a unit descriptor, cumulative equivalent minutes that the T₉₀ temperature was equal to 43°C (CEM43T₉₀), was developed [20] that takes into account both treatment time and the low end of the temperature distribution since T₉₀ is the temperature reached or exceeded by 90% of measured temperature points during a HT fraction. In this way, CEM43T₉₀ represents a volumetric thermal dose descriptor that relates the tissue temperature distribution and time of heating.

The target T₉₀ required for study entry was 40.1°C. This is the same definition of tumour heatability that we used in previous canine trials [7]. To evaluate whether this target T₉₀ could be achieved, temperature descriptors were assessed in real time during the first hyperthermia fraction beginning immediately following power application. Once an instantaneous T₉₀ of 40.1°C was observed, the tumour was randomised to receive either one hyperthermia fraction per week for 5 weeks or three to four hyperthermia fractions per week, also for 5 weeks (16–18 total). Monitoring tumour heatability and performing randomisation during the first hyperthermia fraction eliminated the possibility of entering a non-heatable tumour (there were none) and then allowed for tailoring of the remainder of the first hyperthermia fraction according to whether one fraction per week or three to four fractions per week were desired.

The hypothesis was that the lower fraction-specific temperatures associated with administration of the total thermal dose in 16–18 fractions versus five, would lead to a favourable increase in tumour oxygenation in the 16–18 fraction group compared to five fractions and that the increased oxygenation would lead to decreased tumour volume and increased apoptosis and necrosis in irradiated tumours. Randomisation was stratified by tumour volume (1 − <60 cm³ versus 60–400 cm³) and pretreatment tumour oxygenation (median pO₂ <15 mm Hg versus ≥15 mm Hg) to balance these outcome variables between groups. The cut-off point for tumour volume of 60 cm³ and for median pO₂ of 15 mm Hg was chosen as these were the median values for these variables in a prior canine trial [7].

For the five hyperthermia fraction group, the target T₉₀ was ≥40.5°C, with the duration of each fraction adjusted such that 4–10 CEM43T₉₀ were given per fraction to meet the total thermal dose goal of 20–50 CEM43T₉₀. For the 16–18 hyperthermia fraction group, the target T₉₀ remained 40.1°C, with the duration of each fraction adjusted such that ∼1.5–2 CEM43T₉₀ were given per fraction to meet the total thermal dose goal of 20–50 CEM43T₉₀. Dogs received concurrent daily fractionated radiation therapy of 25 fractions of 2.25 Gy using 6-MV photons for a total dose of 56.25 Gy in 5 weeks. Hyperthermia was always administered prior to irradiation; the inter-treatment interval was approximately 1 h.

Measurements of tumour volume, tumour oxygenation, and diffusion-weighted MRI were repeated 24 h after the first hyperthermia fraction. Tumour biopsies were also repeated 24 h after the first hyperthermia fraction and during the second and third weeks of treatment. Tumour oxygenation was measured 10–12 additional times during treatment and at the end of treatment. Interstitial pressure was measured midway through treatment and at the end of treatment. Diffusion-weighted MRI and tumour biopsies were repeated at the end of treatment.

Tumour biopsies were assessed quantitatively for apoptosis, necrosis, microvascular density (von Willebrand factor, vWF), vascular endothelial growth factor (VEGF), hypoxia inducible factor 1α (HIF-1α), and carbonic anhydrase IX (CA-IX). The initial tumour biopsy was also assessed for tumour grade using the number of mitotic figures per high field as the measure. For all immunohistochemistry, Tru-Cut biopsy samples were paraffin embedded and cut into 5-um sections that were deparaffinised and rehydrated using Citrisolve (Amity International, Barnsley, UK) and graded ethanol. Antigen retrieval was performed on all samples using 10 mM sodium citrate buffer at pH 6.0 heated to 95°C for 20 min. All samples were blocked against secondary antibody background with 10% normal donkey serum in PBS for 1 h and all sections were incubated with primary antibody overnight at 4°C. Endogenous peroxidase was quenched with 3% H₂O₂ for 30 min at room temperature following primary antibody incubation. Following secondary antibody incubation all slides had avidin-biotin complex (Vector Labs ABC Elite avidin-biotin linker kit, PK-6100, Burlingame, CA) applied for 30 min at room temperature. Lastly, DAB substrate (Vector Labs, SK-4100) was applied and sections were incubated for 5 min. Slides were counterstained with haematoxylin for 30 s, dehydrated and mounted.

VEGF

Sections were incubated overnight at 4°C with R & D Systems, ab MAB1603 (Minneapolis, MN), at a concentration of 25 µg/mL (1:20 dilution). The secondary antibody, biotinylated donkey anti-mouse (Jackson Immuno, 715-065-150, West Grove, PA) at a concentration of 1:2000 was applied and incubated for 1 h at room temperature.

CA-IX

Sections were incubated overnight at 4°C with Abcam ab 15086 (Cambridge, MA), at a concentration of 2 µg/mL (1:500 dilution). The secondary antibody, biotinylated donkey anti-rabbit (Jackson Immuno, 715-065-152) at a concentration of 1:2000 was applied and incubated for 1 h at room temperature.

vWF

Sections were incubated overnight at 4°C with DiaPharma SACWF-IG (Franklin, OH), at a concentration of 2.5 µg/mL (1:2000 dilution). The secondary antibody, biotinylated donkey anti-sheep (Jackson Immuno, 715-065-003) at a concentration of 1:2000 was applied and incubated for 1 h at room temperature.

HIF-1α

Sections were incubated overnight at 4°C with Affinity Bioreagents PA1-16601 (Thermo Fisher Scientific, Rockford, IL) at a concentration of 2.5 µg/mL (1:100 dilution). The secondary antibody, biotinylated donkey anti-rabbit (Jackson Immuno, 715-065-152) at a concentration of 1:2000 was applied and incubated for 1 h at room temperature.

Apoptosis, based on activated caspase 3

Sections were incubated overnight at 4°C with R&D Systems AF835, at a concentration of 0.5 µg/mL (1:1000 dilution). The secondary antibody, biotinylated donkey anti-rabbit (Jackson Immuno, 715-065-152) at a concentration of 1:2000 was applied and incubated for 1 h at room temperature.

All sections were anonymised and randomised using a random number generator available at www.random.org. Values were recorded and then slides were un-anonymised.

Assessment criteria for HIF-1α, CAIX, and VEGF

To be considered assessable by immunohistochemistry, tumour sections needed to contain viable tissue of at least one 40× field of view. Any necrotic areas were not assessed with this parameter. The tumour biopsies that contained multiple pieces of tumour had each piece scored individually and those scores averaged. The percentage of area stained and the intensity of the staining were both scored at 10×. Intensity was classified as a 0, 1, 2, or 3. Percentage area was classified as none (0), 1–25% (1), 26–50% (2), 51–75% (3), or 76% (4) or greater. A final immunohistochemistry score was determined by multiplying the intensity by the percentage area of staining.

Assessment criteria for vWF (microvessel density):

Tumour sections were assessed with three 40× fields. Any necrotic areas were not assessed with this parameter. Cells that stained positive were counted and the total number of cells contained in three hot spots was recorded.

Assessment criteria for caspase 3 (apoptosis):

Tumour sections were assessed with five 100× fields. Any necrotic areas were not assessed with this parameter. Cells that stained positive were counted and the total number of cells contained in the five hot spots was recorded.

The primary response end points were necrosis and apoptosis in the tumour as assessed histologically in the tumour biopsies, and percentage tumour volume change at the end of treatment. The effects of temperature and fractionation on water diffusion were also assessed.

Statistical method

Descriptive statistics were used to summarise the baseline tumour information and thermal parameters, tumour oxygenation, interstitial fluid pressure and immunohistochemistry scores. The Wilcoxon rank-sum test or two-sample test were used to study the treatment difference in continuous variables. The chi-square test was applied to examine differences of categorical end points between two treatment groups. Fisher's exact test was conducted to study if there was a non-random association between the categorical end points and the treatment group. Pearson's product moment correlation coefficients and p values to test the significance of correlation were calculated to assess the correlation between two variables.

Results

Thirty-seven dogs were entered; 21 were randomised to one hyperthermia fraction per week and 16 to three to four hyperthermia fractions per week. Twenty-nine of the 37 dogs had diffusion-weighted MRI.

Most tumours in each treatment group were extremity tumours (Table I). There were no significant differences in any of the following pretreatment variables between treatment groups: tumour volume (Table I), tumour oxygenation, interstitial fluid pressure, VEGF, CAIX, HIF, vWF, apoptosis and necrosis. There were more high grade tumours in the three to four fractions per week group (Table I).

Table I.  Distribution of tumour grade, tumour site and tumour volume between treatment groups. Proportionally more high-grade tumours were present in tumours treated with three to four fractions per week.

Variable	Arm		P value
	1 per week	3–4 per week
Tumour grade
Low/intermediate	20 (95.2%)	7 (43.8%)	0.0007*
High	1 (4.8%)	9 (56.2)
Tumour site
Head	1 (4.8%)	2 (12.5.5)	0.23*
Trunk	2 (9.5.5)	4 (25.0%)
Extremity	18 (85.7%)	10 (62.5%)
Tumour volume (cm^3)
Mean	69.2	89.7	0.87
Median	72.3	60.1
Min	11.0	10.0
Max	142.6	360.3

Note: *Fisher's exact p value.

Individual hyperthermia treatments were longer in the three to four fractions per week group and intratumoural temperatures at each fraction were lower, compared with the one fraction per week group (Table II). This was intentional and occurred because of adjustment of the T₉₀ and treatment time to give between 4–10 CEM43T₉₀ per fraction in the one fraction per week group, and between 1.5–2 CEM43T₉₀ per fraction in the three to four fractions per week group so that the total prescribed dose of 20–50 CEM43T₉₀ would be met in both groups despite the large difference in the number of hyperthermia fractions. The total cumulative thermal dose, quantified as CEM43T₉₀ was statistically different between treatment groups but was within the target range (20–50 CEM43T₉₀). Although statistically different, it is doubtful that the small difference in total CEM43T₉₀ values between groups was clinically significant.

Table II.  Thermal parameters as a function of treatment group. Tumour temperatures were higher and duration of each hyperthermia treatment shorter in the one fraction per week group compared to the three to four fractions per week group. The total cumulative thermal dose, quantified as CEM43T₉₀ was slightly higher in the one fraction per week group but was within the target range of 20–50 CEM43T₉₀. It is doubtful that the small absolute difference in total CEM43T₉₀ values between groups is clinically significant.

	Arm		P value
Variable	1 per week (n = 21 dogs)	3–4 per week (n = 16 dogs)
T10 (°C) at first hyperthermia treatment
Mean	44.90	43.37	0.0013
Median	45.13	43.38
T50 (°C) at first hyperthermia treatment
Mean	42.91	41.47	<0.0001
Median	43.10	41.31
T90 (°C) at first hyperthermia treatment
Mean	41.15	40.08	<0.0001
Median	41.02	40.04
Total number of hyperthermia treatments
Mean	4.9	15.4	<0.0001
Median	5.0	16.0
Duration of first hyperthermia treatment (min)
Mean	55.8	82.6	0.009
Median	45.0	89.5
Total CEM43T10 (min)
Mean	1185.7	2109.1	0.044
Median	749.8	1814.0
Total CEM43T50 (min)
Mean	251.7	304.0	0.318
Median	200.5	291.1
Total CEM43T90 (min)
Mean	29.9	24.9	0.007
Median	32.1	24.2
Total duration of hyperthermia (min)
Mean	240	1459	<0.0001
Median	206	1506

The main finding of this fractionation comparison was a statistically greater reduction in tumour volume at the end of treatment in the one fraction per week group compared to the three to four fractions per week group (p = 0.0022) (Figure 1). Additionally, but without respect to treatment group, tumour oxygenation tended to decrease 24 h after the first hyperthermia fraction (Table III). However, there was a weak (correlation coefficients ∼0.4) association between an improvement in oxygenation at 24 h and reduction in tumour volume at the end of treatment, whether tumour oxygenation was based on median pO₂ (p = 0.0146), percentage measured O₂ points ≤2.5 mm Hg (p = 0.0138) (Figure 2) or percentage measured O₂ points ≤5 mm Hg (p = 0.0286), A similar relationship was also found for changes in oxygenation between 7 and 15 days relative to the pretreatment value (data not shown as a relationship is similar to that for 24 h after treatment).

Figure 1. Percentage tumour volume change at end of treatment as a function of treatment group. There was a statistically significant greater reduction in tumour volume at the end of treatment in tumours treated with one fraction per week compared to tumours treated with three to four fractions per week (p = 0.0022).

Figure 2. Median number of measured pO₂ points <2.5 mm Hg (24 h post treatment minus pretreatment) as a function of the percentage change in tumour volume at the end of treatment. Note the direction of the x-axis; decreasing tumour volume proceeds to the right. The pattern is a greater tumour volume reduction in tumours where the median number of measured points <2.5 mm Hg decreases compared to the pretreatment value, i.e. improving oxygenation (p = 0.0138, test for zero correlation; correlation coefficient 0.41).

Table III.  Median descriptors of tumour oxygenation prior to treatment versus 24 h after the first hyperthermia fraction. These data are for all tumours, in both fractionation groups combined. The change in median values is in a direction consistent with reduction in oxygenation, without respect to fractionation group.

Variable	Pretreatment	24 h after first HT treatment	P value
Median pO2	10.5	6.2	0.06
Median % measured pO2 points ≤2.5 mm Hg	22.5	34.2	0.44
Median % measured pO2 points ≤5 mm Hg	33.4	50.0	0.03
Median % measured pO2 points ≤10 mm Hg	48.9	61.8	0.02

Notes: HT, hyperthermia; SD, standard deviation. P value is from signed rank test to compare medians.

At the end of treatment, the lower end of the range of ADC values decreased in the one fraction per week group in comparison to a slight increase in the lower end values in the three to four fractions per week group (Figure 3). In other words, if one looks at the change in water diffusion in an individual tumour in the one fraction per week group, regions where water diffusion was more restricted before treatment were characterised by greater decreases in water diffusion following treatment than regions where water diffusion was less restricted before treatment (Figure 4).

Figure 3. Difference in median percentile values for the apparent diffusion coefficient of water (ADC), post-treatment relative to pretreatment, as a function of the individual percentile as a function of treatment group. The greatest decrease occurred in the low end of the diffusion range in dogs receiving one fraction per week (0.0023 ≤ p ≤ 0.4322). 5HT, one hyperthermia fraction per week; 20 HT, three to four hyperthermia fractions per week.

Figure 4. Percentile distribution of apparent diffusion coefficient (ADC) values in an individual tumour before treatment and at the end of treatment. In this subject, treatment led to a generalised decrease in water diffusion but regions where water diffusion was more restricted prior to treatment (low ADC values) were characterised by greater decreases. Every 1000th point is shown. Fewer points are present after treatment because the tumour was smaller. Pre, before treatment.

The direction of change of HIF-1α and CA-IX immunoreactivity 24 h after the first hyperthermia fraction were related directly. Of 19 dogs with a reduction in CA-IX immunoreactivity at 24 h, 17 (89%) also had reduction in HIF-1α immunoreactivity. Also, of 19 other dogs with an increase in CA-IX immunoreactivity at 24 h, 14 (74%) also had an increase in HIF-1α immunoreactivity (p < 0.0001, Fisher's exact test). The direction of change of CA-IX was also related to thermal dose administered during the first hyperthermia fraction. Median CEM43T₉₀ for the first hyperthermia fraction in dogs with a decrease in CA-IX at 24 h was 4.6 min versus 1.9 min in dogs with an increase in CA-IX at 24 h.

There were no treatment or temperature effects on changes in interstitial fluid pressure, VEGF, vWF, apoptosis or necrosis (data not shown).

Discussion

Based on percentage tumour volume change at the end of treatment, administering a prescribed hyperthermia dose using one fraction per week was more effective than when administering it in three to four fractions per week (Figure 1). As the temperatures were higher at each fraction in the one fraction per week group, this greater volume reduction may have been due to greater cytotoxicity at each fraction. Alternatively, thermotolerance developing in the three to four fractions per week group may have diminished the efficacy of that prescription, and/or temperatures might not have been high enough to cause clinically significant cytotoxicity.

Percentage change in tumour volume is not the most robust end point to use to assess treatment efficacy in a solid tumour. The time to an event, such as local recurrence, would have been superior but survival analyses are difficult to implement in a canine tumour model due to the large number of subjects required to detect a clinically significant difference between treatment groups, unless the difference is very large. Although we previously completed a time-to-local-control analysis in a canine sarcoma model, the on-study time was approximately 7 years [7]; this is too long to complete exploratory trials. Additionally, using only radiation and hyperthermia to treat canine soft tissue sarcomas is no longer acceptable clinically due to the proven value of surgery for this tumour type. Nevertheless, the use of tumour volume change as a surrogate for local control has been documented as a valid end point [21, [22]]. Importantly, a volume reduction of >50% for human sarcomas is highly predictive of a good pathological response [22]. Approximately 75% of the canine sarcomas in this trial treated with one hyperthermia fraction per week had a volume reduction of at least 50% compared to very few sarcomas receiving three to four hyperthermia fractions per week, as seen in Figure 1. Thus, applying the guidelines of Roberge [22], most sarcomas receiving one fraction per week would be expected to have a good pathological response compared to few to no sarcomas treated with three to four fractions per week. This greater effect of the one fraction per week on volume reduction is an important finding and one that should be considered when future clinical trials are designed.

We hypothesised that administration of a defined thermal dose in three to four fractions per week would lead to improved tumour oxygenation but there was no treatment group effect on changes in tumour oxygenation, based on invasive probe measurements. Invasive oxygen measurements have been considered to be the gold standard in many clinical trials where tumour oxygenation was being quantified and results have led to the identification of important clinical principles. Regardless, invasive measurements are subject to sampling error and may also perturb perfusion due to the invasiveness of the procedure, especially when conducted longitudinally over 5 weeks, as herein. Also, in human trials, serial invasive measures are not likely to be feasible due to patient comfort and compliance issues. To solve these problems, non-invasive functional imaging of oxygenation using PET, with radiopharmaceuticals such as ¹⁸F-EF5 [23] or ⁶²Cu-ATSM [24] would be better tolerated and will also allow quantification of tumour oxygenation volumetrically, such that it can be compared on a 3D voxel basis to other parameters measured volumetrically, such as perfusion and temperature.

Although we did not find a significant effect of treatment group on the change in tumour oxygenation, we did find a significant association between improvement in tumour oxygenation at 24 h after the first hyperthermia fraction, and also improvement in oxygenation between 7 and 15 days (data not shown), and percentage reduction in tumour volume at the end of treatment (Figure 2). This is logical in that tumours with improved oxygenation would be expected to be more responsive to radiation. A significant treatment group effect may not have been found due to the variation in oxygenation changes within each group and the association only became apparent when looking specifically at individual oxygen parameters. This result suggests that quantification of the change in tumour oxygenation 24 h after the first hyperthermia fraction may be a reliable end point to use to predict the success of a fractionated hyperthermia prescription when combined with radiation, regardless of the specific hyperthermia fractionation details.

Developed originally as a way to characterise acute thrombotic brain infarction, diffusion weighted imaging is also receiving attention as a predictor of tumour response [25]. As water diffusion in tumours is restricted by cellular membranes and macromolecular structures, cytotoxic treatment can lead to loss of cell membrane integrity, which can be detected as an increase in water diffusion values for the tumour. This will be reflected as an increase in the apparent diffusion coefficient (ADC) of water in diffusion weighted imaging [25]. The assessment of the ADC is best done at the percentile level rather than simply looking at median or mean values because changes in the distribution pattern can occur without a change in the mean value [26]. This is similar in concept to assessing tumour temperature percentiles, i.e. the T90, rather than the median or mean. We found that the lower percentiles of the ADC values were decreased in the one fraction per week group at the end of treatment (∼5 weeks) in comparison to a slight increase in the lower end percentiles of the ADC values in the three to four fractions per week group (Figure 3). In other words, in tumours treated with one fraction per week (higher temperatures per fraction), regions where water diffusion was more restricted before treatment were characterised by greater decreases in water diffusion following treatment than regions where water diffusion was less restricted before treatment (Figure 4). This is an interesting observation, but based on a greater tumour volume decrease in dogs receiving one hyperthermia fraction per week, the finding of reduced water diffusion goes against the conventional interpretation of the predictive value of the ADC where favourable tumour responses are usually associated with increases in the ADC. What specifically caused the greater reduction in water diffusion in tumours treated with fewer fractions of higher temperatures is unknown. With treatments such as radiation therapy or chemotherapy, an increase in the ADC is generally associated with greater cytotoxicity as a result of cell killing leading to less diffusion restriction. We observed the opposite; a reduction in the ADC at the end of therapy in tumours treated with more effective hyperthermia fractionation (one per week) in terms of observed volume reduction. However, there is very little known about the relationship between the direction of change or temporal kinetics of change of the ADC and tumour response in hyperthermia trials, or the effect of tissue geometry on the value of the ADC. In uterine fibroids treated with radiofrequency ablation, the response of the ADC parameters was variable, increasing in some tumours and decreasing in others, with no predictive power [27]. In normal canine prostate treated with radiofrequency ablation, the ADC value decreased immediately after treatment but increased subsequently as prostate recovered [28]. Our finding of greater restriction in water diffusion at the end of treatment, in tumours with reduced volume, may be related to a condensation of stroma, producing a water restriction barrier, rather than to increased cellularity as might be concluded from application of conventional thinking regarding the change in the ADC value versus cellularity. It is important to re-emphasise that, in addition to pretreatment, we applied diffusion weighted MRI to quantify ADC at 24 h after the first hyperthermia fraction and at the end of therapy, approximately 5 weeks later. These times are probably too soon and too late, respectively, to quantify water diffusion as a reflection of effectiveness of tumour therapy. Further work is needed with diffusion weighted imaging being performed more often and closer to the beginning of therapy but longer than 24 h after the first fraction, before significant volume change from the radiation/hyperthermia becomes widespread.

Unfortunately, there were more high-grade tumours in the three to four fractions per week group, which may have contributed to the smaller treatment effect, in terms of tumour volume change, in this group (Figure 1). We did not stratify randomisation by tumour grade as there is a limit to the number of stratification variables that can be employed in studies where the overall number of subjects is relatively small. We thought it more important to balance the primary study end point variables, i.e. tumour volume and tumour oxygenation. Regardless, there was no association between tumour volume change at the end of therapy and tumour grade. Intermediate/low-grade tumours underwent 52.4% volume reduction compared to 37.8% volume reduction in high-grade tumours (p = 0.11, ANOVA), suggesting that tumour grade was not a highly influential variable. Also, one might expect a more dramatic volume reduction in high grade tumours due to the higher mitotic index, leading to greater responsiveness, as has been observed in chemotherapy trials [29]. In that scenario, having more high-grade tumours in a group might actually inflate the treatment effect. However, it is just not possible to know how the disparity of tumour grade affected the results. Regardless, the available results supported one hyperthermia fraction per week being more effective in terms of tumour volume reduction compared to three to four hyperthermia fractions per week, when a constant total thermal dose was given. In future trials, balancing tumour grade between groups should be considered.

Finally, we observed a temperature dependent decrease in CA-IX immunoreactivity, and a direct association between changes in CA-IX and HIF-1α immunoreactivity. The expected association between hyperthermia and the directional change in HIF-1α immunoreactivity has not been characterised completely. An inverse association between temperature and HIF-1α immunoreactivity was documented in macrophages [30], but a direct association between temperature and HIF-1α immunoreactivity observed in tumour cells [31, [32]] and murine tumours [33]. As such, we would have expected a direct association between temperature and either HIF-1α immunoreactivity or one of its downstream products, but we observed a decrease. The reason for this unexpected result is not known. This may be a temperature-dependent effect, which has not been characterised completely, especially in the situation of highly heterogeneous temperatures, as occurred in these canine tumours. Finally, it might have been expected to see increased VEGF immunoreactivity as a consequence of the up-regulation of HIF-1α. We did not observe this and this may be due to sampling error, which undoubtedly affected the quantification of all histologic parameters, again pointing to the value of non-invasive volumetric determination of tumour parameters in future studies where tumour physiology is being assessed.

We found no significant change in interstitial fluid pressure, VEGF, vWF, apoptosis or necrosis. This should not be taken as evidence that these end points are not important in terms of tumour response to hyperthermia. Perhaps the sensitivity of our sampling and quantification methods was not adequate to detect the changes that occurred.

Conclusion

In conclusion, based on changes in tumour volume, our evaluation of hyperthermia fractionation in canine sarcomas suggests that administration of a large number of relatively low-temperature hyperthermia fractions on a multiple fraction per week schedule is not associated with improved tumour response. We found greater tumour volume reduction in tumours treated with one fraction per week. Importantly, we are not implying that one fraction per week is optimal, only that three to four fractions of relatively low temperature administration does not seem to confer a benefit. The value of administering a large number of hyperthermia fractions has been debated for years, with the suggestion that the lower temperatures would lead to greater increases in tumour oxygenation. Although this study was undoubtedly influenced by the invasive methods used to characterise the tumour, and possibly by the difference in tumour grade between groups, we found no advantage of a highly fractionated prescription. In fact, there was evidence that a coarsely fractionated prescription of one fraction per week was superior. This needs to be tested in a spontaneous tumour where the physiological end points are quantified volumetrically and not invasively as herein, and all influential variables are balanced between groups by means of stratification, or counteracted by a large sample size. Only through further prospective investigation where thermal dose is defined a priori and administered strictly according to protocol will optimum hyperthermia fractionation be defined. Defining optimal hyperthermia fractionation will likely be paramount in re-establishing the value of hyperthermia for cancer treatment in the USA.

Declaration of interest

This work was supported by grant number P01 CA42745 from the US National Institutes of Health. The authors alone are responsible for the content and writing of the paper.