Utility of functional imaging in prediction or assessment of treatment response and prognosis following thermotherapy

Abstract

The purpose of this review is to examine the roles that functional imaging may play in prediction of treatment response and determination of overall prognosis in patients who are enrolled in thermotherapy trials, either in combination with radiotherapy, chemotherapy or both. Most of the historical work that has been done in this field has focused on magnetic resonance imaging/magnetic resonance spectroscopy (MRI/MRS) methods, so the emphasis will be there, although some discussion of the role that positron emission tomography (PET) might play will also be examined. New optical technologies also hold promise for obtaining low cost, yet valuable physiological data from optically accessible sites. The review is organised by traditional outcome parameters: local response, local control and progression-free or overall survival. Included in the review is a discussion of future directions for this type of translational work.

• MRI
• PET
• optical spectroscopy
• prognosis, functional imaging
• hyperthermia

Introduction

In this era of personalised medicine, much effort is being directed to the development of methods that can assist in the selection of the best therapeutic options for individual cancer patients. The drive toward individualisation of treatment has led to the publication of literally thousands of papers on this subject and the generation of journals that are specifically dedicated toward this goal. Examples include: The Open Biomarkers Journal–http://www.bentham.org/open/tobiomj/; Biomarker Insights–http://www.la-press.com/biomarker-insights-journal-j4; and Disease Markers–http://www.iospress.nl/loadtop/load.php?isbn=02780240.

One of the most intensely investigated methods for individualising cancer treatment has been in the characterisation of genomic markers of biologic diversity. A myriad of host- and tumour-specific variations in gene expression occurs in cancer, which provides a rationale for believing that individualisation of treatment is possible [1], [2]. Indeed, genomic profiling in particular holds great promise for prediction of chemotherapy response in numerous tumour types, including breast cancer [3], [4], prostate cancer [5] and non-small-cell lung cancer [6], [7]. In the field of hyperthermia, there has been limited activity in this area with only one study being reported thus far, wherein a genomic profile was derived for prediction of persistence of positive lymph nodes following neoadjuvant thermochemotherapy in locally advanced breast cancer [8].

In spite of the enthusiasm for genomic profiling or other types of molecular and cellular biomarkers, one cannot ignore the physiological microenvironment in tumours, particularly when considering use of hyperthermia. With therapeutic hyperthermia, perhaps more than any other cancer treatment modality, the physiological microenvironment can affect treatment delivery, can influence treatment outcome, and can be altered by the treatment both positively and negatively. For example, tumour perfusion will be inversely related to the ability to heat a tumour using external methods, heat-induced vasculopathy may lead to up-regulation of oxygen-sensitive proteins that could increase the biological aggressiveness of the tumour, hyperthermia might increase tumour oxygenation leading to enhanced radiosensitivity, and acute changes in pH could increase the chance of thermal cytotoxicity.

In this review we focus on examining how functional imaging parameters that reflect the tumour microenvironment can be used to predict treatment outcome and overall survival in thermotherapy trials. Endpoints that have been assessed using functional imaging include tumour response, duration of local tumour control and progression free and overall survival.

Overview of molecular imaging methods

Magnetic resonance imaging, dynamic contrast enhanced MRI and magnetic resonance spectroscopy

MRI is based on the principle of nuclear magnetic resonance, where nuclei in an external magnetic field selectively absorb and then release energy unique to those nuclei. The released energy can be used to create an anatomic image that is characterised by outstanding contrast resolution. The outstanding contrast resolution and lack of a need for ionising radiation have led to MRI being widely used for tumour detection and staging and for assessing response to treatment. Though the signals used to create a magnetic resonance image can be affected by the microenvironment, MRI is essentially an anatomic imaging modality. However, modifications of basic MRI techniques can be used to gain more functional information.

In DCE-MRI, the kinetics of intratumoural water soluble contrast medium concentration, i.e. signal, as a function of time after intravenous injection, allows physiological parameters related to tissue perfusion/permeability to be evaluated. Methods of evaluation include visual inspection of data in movie format, inspection of graphs of signal intensity versus time, empirical institution dependent measurements and pharmokinetic modelling using multi-compartment analysis [9]. There are advantages and disadvantages to each of these analytical methods, but overall the lack of a universally accepted analytical method has likely impeded understanding the full potential of this non-invasive imaging method for staging patients and predicting outcome [10]. Nevertheless, the non-invasive nature of DCE-MRI and the feasibility of follow-up studies are attributes that make DCE-MRI attractive for assessing tumour perfusion in clinical trials.

MRS differs from MRI in that in the latter frequency is used for spatial localisation, while in the former frequency is used for both for spatial localisation and chemical identification [11]. Thus, in MRS, specific chemicals can be detected based on their frequency signature. Protons are usually used for spectroscopy due to their abundance and high magnetic sensitivity, though special techniques allow other nuclei, such as phosphorus or fluorine, to be evaluated as well. Due to the pH sensitivity of the frequency signature of certain molecules, MRS can also be used for non-invasive measurement of intracellular pH [12]. Non-invasive assessment of chemical spectra can provide unique information regarding the evaluation of tumour behaviour and possibly response to treatment.

Positron emission tomography

PET imaging can provide qualitative and quantitative metabolic information using molecular probes labelled with positron-emitting radionuclides, such as ¹⁸F, ¹¹C, ¹⁵O, ^60,62,64Cu, ¹²⁴I. A positron emitted by a radiotracer travels only a small distance before it encounters an electron, at which point an annihilation reaction occurs where matter is converted to energy, and two 511 keV photons are emitted in opposite directions. The PET scanner detects both these photons using a cylindrical array of detectors. The fact that the two photons are produced simultaneously and travel at 180° to one another is used to further localise each event. Most PET scans are now performed on hybrid PET/CT scanners. This allows the anatomic information from CT and metabolic information from PET to be obtained in a single imaging session. The CT scan is also used to provide attenuation-correction of the PET images, enabling activity within tumours to be accurately measured. Using PET, radiotracer uptake within tumours is quantified routinely by measuring standardised uptake value (SUV), which normalises the radiotracer activity to body weight and administered radiopharmaceutical dose.

At present the most commonly used PET radiotracer is 2-deoxy-2-[¹⁸F]fluoro-D-glucose (FDG). FDG is a glucose analog that is actively transported into cells where it is trapped and not further metabolised. Malignant cells use more glucose than normal cells (Warburg effect), and often have higher glycolytic metabolism related to tumour hypoxia. In general, aggressive malignancies exhibit higher accumulation of FDG compared to lower-grade tumours, and the degree of FDG activity may provide prognostic information. Malignant tumours responding to chemotherapy are typically characterised by a measurable reduction in FDG uptake on PET scans before becoming measurably smaller on CT or MRI scans, and therefore may be predictive of early response to therapy [13].

While FDG is currently the dominant radiotracer used in PET imaging, other tracers are being developed which may be useful in determining response to therapy but are not yet available for routine clinical use; these are summarised in a recent review by Larson [14]. Examples include ¹⁸F-fluourothymidine (FLT) as a marker for cell proliferation and hormonal markers such as ¹⁸F-fluouroestradiol (FES). Several PET tracers are being studied as surrogate markers for hypoxia. These include 2-nitroimidazole agents such as ¹⁸F-fluoromisonidazole (FMISO), ¹⁸F-fluoroetanidazole (FETA), ¹⁸F-fluoroazomycin-arabinofuranoside (FAZA), and ¹⁸F-fluoroalkyl acetamide derivatives (EF1, EF3, and EF5) [15], [16]. These agents diffuse freely throughout the body after injection so that delivery to target areas is independent of blood flow. Under conditions of hypoxia, the tracer is reduced and irreversibly trapped intracellularly. Copper-labelled dithiosemicarbazones are also reduced and retained in hypoxic tissues. Cu-ATSM (diacetyl-bis(N⁴-methylthiosemicarbazone) can be labelled with several positron-emitting copper isotopes, including ⁶⁰Cu (t_1/2 = 23 min), ⁶²Cu (t_1/2 = 9.7 min), and ⁶⁴Cu (t_1/2 = 12.7 h). Cu-ATSM accumulates more rapidly in hypoxic tissues compared to FMISO, but is more sensitive to blood flow effects and tumour type [15], [17]. In the future, PET imaging with hypoxia markers may be useful for characterising tumours, and for evaluating response to chemoradiation and thermal therapy.

Optical imaging

Optical spectroscopy involves the interrogation of tissue optical properties to gain insight into a variety of functional and structural features. There are a wide range of potential sources of optical contrast in tissue, but the properties focused on in this review involve the absorption, scattering, and fluorescence properties of tissue. Such measurements have a variety of advantages, including safety (non-ionising), low-cost, quantitative, non-destructive, and wide sensitivity to a range of functional parameters. Optical measurements are commonly made using a fibre-optic probe placed in contact with the tissue of interest, or using a non-contact imaging system [18]. Light enters the tissue and undergoes absorption, scattering, and or fluorescence. Exiting photons can then be collected and recorded with an optical detector. The penetration depth of light ranges from a few hundred microns in the UV, up to several cm in the near infrared, making this technique applicable to superficial tumours as well as sites such as the intact breast using NIR wavelengths [19]. Measurement via a biopsy needle or endoscope enables measurement of more deeply seated tumours [18].

Two types of measurements possible using this type of set-up will be discussed: diffuse reflectance and fluorescence spectroscopy. Diffuse reflectance involves illuminating the tissue and recording the backscattered light as a function of wavelength. This is dependent on the wavelength-dependent absorption and scattering properties of the tissue. Fluorescence spectra are recorded by illuminating the tissue with a given wavelength of light, and recording the intensity emitted at a longer wavelength. This relies on tissue fluorophores to absorb the incident photon and emit a fluorescent photon at a longer wavelength. This measurement is sensitive to the absorption, scattering, and fluorescence properties of the tissue. Tissue is highly scattering, with scattering tending to decrease at longer wavelengths, and emitted photons are typically multiply scattered. Absorption in tissue in the visible wavelength range is dominated by haemoglobin, and tends to increase towards the ultraviolet wavelength range. This turbidity makes determination of the underlying optical properties non-trivial due to the complex interactions of light with tissue, but a variety of quantitative models of light–tissue interaction have been developed that are capable of extracting the absorption, scattering, and fluorescence properties from measured tissue spectra [20].

The functional parameters to which optical spectroscopy is sensitive include: (1) total haemoglobin content of tissue which is related to the vascular fraction of the tissue, (2) the haemoglobin oxygen saturation, which is related to vascular oxygenation, (3) scattering properties, which are affected by tissue morphology including cellular and extracellular matrix density, and (4) intrinsic sources of fluorescence, including the electron carriers NADH and FAD [18]. This enables optical spectroscopy to be intrinsically sensitive to a wide range of functional and morphological properties, many of which are known to be affected by thermotherapy.

The turbidity of tissue to optical radiation serves to limit the effective penetration depth and resolution at which tissue can be characterised at depth. The effective penetration depth ranges from several hundred microns or less in the UV, up to several cm in the near infrared [18]. This limits the types of measurements that can be made in some clinical settings, where the site of interest is not superficial. Several solutions to this problem have been used, including limiting the measurement to NIR wavelengths, where the penetration depth is greater but fewer sources of contrast are available, or using endoscopic or needle-based measurement systems to access deep tissue [18].

Prediction and assessment of tumour response

The change in ATP content within tumours has been used to predict response rates. Ohtsubo described a linear relationship between decreases in tumour cell ATP content and decreases in cell survival when cells were heated in vitro using a range of temperatures [21]. This provides a rationale for considering change in tumour ATP content as a surrogate for cell viability. Vaupel [22] used 31-P MRS to measure changes in 31-P metabolites in response to hyperthermia. Vaupel used graded thermal doses by changing the time of heating at 43.5°C and found that inorganic phosphate (Pi) increased with heating time, whereas all ATP peaks declined [22].

Sijens et al. found that the ATP/Pi ratio in a murine mammary carcinoma increased slightly 18 h following 42°C heating for 15 min [23]. For higher temperatures, however, there was a temperature-dependent decline in the ATP/Pi ratio. At 45°C the ratio decreased by more than a factor of 10 (Figure 1a). In a thermoradiotherapy trial in sponteneous canine soft tissue sarcomas, 31-P MRS was performed prior to and 24 h after the first hyperthermia treatment [24]. When ATP/Pi was plotted as a function of the median temperature achieved during hyperthermia, the temperature dependence of the ATP/Pi ratio was strikingly similar to that found in the murine mammary adenocarcinoma (Figure 1b). In a parallel trial in human soft tissue sarcomas in which thermoradiotherapy was administered prior to surgical resection, a decrease in adenosine triphosphate/phosphomonoester (ATP/PME) after the first hyperthermia treatment was associated with more necrosis in resected tumours [24] (Figure 1c). Thus, in all of these models, the decrease in ATP observed at higher temperatures was consistent with increased thermal cytotoxicity and this quantification of ATP using MRS may have value as a measure of thermal treatment efficacy.

Figure 1. Changes in ATP/Pi, as measured from 31-P MRS, are temperature dependent. (a) Data from mouse mammary carcinoma grown in the flank measured 18 h after local heating for 15 min at the designated temperature. Figure reproduced with permission from the author and publisher [23]. (b) Data from canine soft tissue sarcomas. Dogs were enrolled on a clinical trial in which hyperthermia was combined with radiation therapy. 31-P MRS studies were performed prior to and 24 h after the first hyperthermia treatment. The change is the difference between these two time points. Data reproduced with permission from the author and publisher [24]. T50 = median temperature. (c) Relationship between change in ATP/PME (pre vs. post first heat treatment) and probability for pathologic CR in human high grade soft tissue sarcomas, based on change in signal between the pre-treatment study and a study performed 24 h after the first hyperthermia treatment [24]. Reproduced with permission from the author and publisher. These data are consistent with the concept that a reduction in ATP after hyperthermia treatment is related to cell killing by hyperthermia.

DCE-MRI has been used to both predict and assess effects of thermal therapy on tumour response. In human soft tissue sarcomas treated neoadjuvantly with thermochemotherapy there was an association between loss of contrast medium uptake and extent of necrosis [25]. Craciunescu et al. developed a semi-quantitative scoring scheme to describe the pattern and kinetics of contrast enhancement prior to the onset of treatment in 20 patients with locally advanced breast cancer [26]. The scoring system has three components: the first is related to morphology and comes from whether the shape of the enhancement pattern as described by the MR parametric maps is centrifugal or centripetal. The second and third components are physiological, one related to tumour vascularity/permeability and the other to tumour extracellular extravascular space, as defined by the wash-in and wash-out parameters (Figure 2). Tumours with a higher score were more likely to respond clinically compared to tumours with a lower score. Low scores were in tumours that likely had relatively poor perfusion. Although based on a very low number of patients, the specificity and sensitivity were 78% and 91% respectively (p = 0.002).

Figure 2. DCE-MRI perfusion patterns in locally advanced breast cancer [26]. Left panel shows a centrifugal pattern, with central contrast medium filling seen in the wash-in and wash-out parameters (top vs. bottom). Right panel is a tumour exhibiting a centripetal pattern. Data reproduced with permission from the publisher and author.

Assessing the effect of hyperthermia on tumour metabolism through PET imaging of glucose analogues is a powerful way to evaluate efficacy. Progressive loss of uptake of FDG as assessed with PET imaging has been found to be related to the amount of necrosis found in excised high grade soft tissue sarcomas that had been treated preoperatively with thermoradiotherapy [27], [28]. Westerterp et al. performed serial FDG PET studies in 17 patients with oesophageal cancer who were treated preoperatively with paclitaxel, cisplatin, radiotherapy and hyperthermia [29]. Patients with greater decreases in FDG uptake after two weeks of treatment had more necrosis in the resected tumour specimen. The positive and negative predictive values were both 75% in this limited series (Figure 3). In an additional report, reduction in FDG PET uptake was a good indicator of pathological response in 20 rectal cancer patients treated with radiotherapy, 5-Fluorouracil (5FU) and hyperthermia [30]. In this study, the second PET imaging study was done after therapy was completed, so it was not tested as a predictor of response.

Figure 3. Receiver operating curve for prediction of pathological CR rate following thermochemoradiotherapy, using change in FDG PET uptake after two weeks of treatment [28]. The receiver operating curve assesses the true positive rate (Y-axis) as a function of the false positive rate (X-axis). An ideal test would yield a true positive rate of 1, and a false positive rate of 0. Data reproduced with permission from the author and publisher.

In summary, there are several methods that appear to reflect the degree of direct cytotoxicity elicited from thermotherapy trials that integrate radiation and/or chemotherapy. Early changes in these parameters may allow accurate prediction of the response at the end of therapy, or at the time of surgical resection. In one study a single measurement of pretreatment tumour perfusion characteristics was predictive of response [26]. In most other studies, a pretreatment determination was compared to a second observation at some point early in the treatment course. Though multiple imaging sessions is a strength of non-invasive tumour imaging, the need for two measurements can occasionally be problematic, as discussed later.

Prediction of local tumour control

Local tumour control is difficult to quantify because of the issue of competing risks. In assessing local control as the primary endpoint there is also usually a risk for development of metastases or death, prior to local failure [31]. These events occurring prior to local failure, the competing risks can be accounted for by censoring in the statistical analysis, but censoring can lead to inaccurate and misleading estimates of the time to local failure [31]. Nevertheless, there are parameters that have allowed prediction of local tumour control, particularly when assessing treatment regimens that include radiation therapy. A prime example is the degree of tumour hypoxia. Tumour hypoxia has been measured most often using invasive oxygen electrodes, but recently non-invasive measurement of tumour hypoxia using PET imaging has been investigated and, similar to invasive measurements, results from PET imaging suggest that hypoxic tumours are more likely to undergo local tumour recurrence after radiotherapy [32], [33]. However, using the same PET imaging approach, others have not found an association between hypoxia and local tumour control following chemoradiation therapy [34]. In soft tissue sarcomas, tumour hypoxia assessed using invasive oxygen electrodes was predictive of metastasis-free survival but not local tumour control following pre-operative thermoradiotherapy and surgical resection [35]. Thus far, PET imaging of hypoxia has not been assessed in a thermoradiotherapy trial where local control was the endpoint.

In most studies where invasive measures of tumour hypoxia have been used, only pretreatment measurements were made, or a small number of measurements were made early in the treatment course. In a study of tumour oxygenation in canine sarcomas where measurements were made throughout a course of thermoradiotherapy, it was apparent that fluctuations in oxygen throughout treatment were more pronounced than expected [36]. This emphasises the value of non-invasive measurement of tumour oxygenation at multiple times in thermoradiotherapy trials so that a more complete understanding of the kinetics of this important microenvironmental factor can be achieved.

MRS has also been used to assess local tumour control in thermoradiotherapy trials. In 29 canine soft tissue sarcomas, the ratio of phosphodiester to phosphocreatine (PDE/PCr) was associated with probability of local tumour control following thermoradiotherapy [37]. The biological significance of the PDE/PCr ratio is difficult to interpret and this observation may have been spurious since a relatively large number of potentially predictive factors were assessed. In this same canine study, however, there was a positive association between the area under the time-intensity contrast enhancement curve and local tumour control. This suggests that the probability of local control was better in more highly perfused tumours. This is similar to that observed in the locally advanced breast cancer study in humans, where more poorly perfused tumours were less likely to respond to thermochemotherapy [26].

Progression-free and overall survival

In several relatively small clinical trials, tumour lactate was quantified using bioluminescence imaging of tumour biopsies and high lactate levels were associated with decreased metastasis-free survival and decreased overall survival. These trials included patients with head and neck [38], cervix [39] and colorectal cancer [40]. There is biological rationale for this observation, as the primary lactate transporters are chaperoned to the cell surface by EMMPRIN (CD147), a protein that is linked to breakdown of extracellular matrix and invasion [41]. MCT-1 and EMMPRIN expression levels and trafficking to the cell membrane are up-regulated upon exposure to elevated levels of lactate [42]. Upon reaching the cell surface, EMMPRIN is shed in vesicles into the extracellular space, where it activates matrix metalloproteinases [43]. Thus, the link between tissue lactate concentrations and more aggressive disease may occur as a result of the concomitant recruitment of EMMPRIN and MCT-1 to the cell surface in response to elevated lactate.

Lactate can be measured using MRS and associations between tumour lactate and survival after radiotherapy of malignant gliomas [44] and non-small-cell lung cancer [45] have been identified. However, there is no information on the importance of lactate in predicting survival in thermotherapy trials.

Extracellular pH (pHe) has been measured using invasive needle electrodes in canine sarcomas treated with thermoradiotherapy [37]. Metastasis-free survival was longer in animals with tumours that had pretreatment pHe values greater than 7 than if the tumour pHe was less than 7 (Figure 4). These less acidic tumours may have had better perfusion. This is also supported by the finding of longer metastasis-free survival in animals with tumours having rapid contrast medium wash-in and wash-out as assessed using DCE-MRI [46] (Figure 5). The results of these two studies provide evidence of a link between pHe and perfusion in influencing treatment outcome in thermo-radiotherapy trials. Further work is needed to assess this important concept.

Figure 4. Probability of metastasis-free survival as a function of time for dogs with tumours having extracellular pH > 7 vs. dogs with tumours having extracellular pH < 7 [36]). Dogs with tumours having a more alkaline extracellular pH had longer metastasis-free survival. Figure reproduced with permission from the author and publisher.

Figure 5. Metastasis-free survival in dogs with tumours dichotomised by contrast medium wash-in rate (a) and contrast medium wash-out rate (b), as determined from DCE-MRI [45]. Metastasis-free survival was longer in dogs with higher wash-in and wash-out values, likely reflecting tumours with better perfusion. The Kaplan Meier curves shown compare the survival for animals, grouped above and below the median for the whole population. Wash-in value is a rate constant derived from the initial slope of the dynamic contrast enhanced image set and is influenced most strongly by the perfusion rate in the tumour [10]. The wash-out value is derived from the terminal slope of the image set and reflects the rate of contrast clearance from the tumour. This is also influenced by perfusion, but is also reflective of the extracellular volume and permeability of the microvasculature. Figure reproduced with permission from the author and publisher.

Phosphorus MRS has been used to predict survival. In humans with high grade soft tissue sarcomas treated preoperatively with thermoradiotherapy, the PME/PDE ratio was predictive of progression-free and overall survival [47]. Differences in this ratio may likely reflect the rate of membrane turnover, which has been hypothesised to be related to tumour cell motility [48]. In a canine trial, the PDE/ATP ratio was found to be of significance with regard to metastasis-free survival. In that trial, subjects with tumours characterised by a high PDE/ATP ratio had significantly longer metastasis-free survival [37]. Given the relationship of elevated PDE with regard to membrane degradation, tumours with a low PDE/ATP ratio may have had increased susceptibility to metastasis.

In the canine and human sarcomas trials, other 31-P MRS parameters were examined for an association with metastasis-free and overall survival, including those that predicted for response, such as ATP/Pi. No significant correlation was between these parameters and outcome was identified in either trial [37], [47].

Optical spectroscopy–a promising new molecular imaging tool

To date, optical spectroscopy has been used primarily for tumour diagnosis, as opposed to assessing prognosis. However, the inherent optical properties that can be derived may be related to prognosis as well. Relevant parameters include total haemoglobin, which is related to vascular volume, haemoglobin saturation and the redox ratio, which are related to hypoxia, and lipid absorption spectra, which are related to lipid content and type. In a small clinical trial in breast cancer, near infrared spectral analysis allowed identification of malignant tumours with 100% specificity and sensitivity [49]. Diffuse optical spectroscopy has also been combined with diffuse correlation spectroscopy, which measures perfusion, to show that these spectroscopic parameters change within a few days after neoadjuvant chemotherapy treatment. It is possible that changes in these parameters early in the course of therapy may be useful in assessing tumour response [50]. Similarly, Brown et al. have reported a significant decrease in the total haemoglobin saturation in malignant breast tumours, as compared with normal or benign lesions [18]. They also report a significant positive correlation between haemoglobin saturation and total haemoglobin content. Finally, they found a significant positive correlation between HER2/neu status and both haemoglobin saturation and total haemoglobin content.

At the preclinical level, we have begun to perform studies to validate optical spectroscopic data against other validated physiologic methods. For example, we found an excellent correlation between changes in haemoglobin saturation and invasive oxygen measurements in flank tumours of mice that are switched from breathing room air to carbogen [51]. We also identified a positive correlation between deoxyhaemoglobin concentration measured with optical spectroscopy and hypoxic fraction assessed with pimonidazole hypoxia marker binding [52]. Additionally, we found a positive correlation between an optical parameter derived from Monte Carlo modelling of diffuse optical reflectance, the backscatter factor (µs), and the histological estimate of percentage necrosis. The primary advantage of optical spectroscopy is that it enables longitudinal monitoring of physiological information, making it ideal to identify transient changes induced by therapy. Figure 6 shows such an example in animals treated with MTD doxorubicin (10 mg/kg i.v.). It can be seen that there is a transient increase in haemoglobin saturation peaking 10 days post-treatment, possibly due to a reduction in oxygen consumption due to tumour cell killing, that may be difficult to pick up using a more limited selection of time points available using immunohistochemistry or other modalities. These results suggest strongly that optical spectroscopy may be a useful tool for monitoring treatment responses that involve thermotherapy. To date, however, optical spectroscopic methods have not been reported in conjunction with thermotherapy studies, either at the preclinical or clinical levels.

Figure 6. Haemoglobin saturation is plotted as a function of time after treatment with MTD doxorubicin. The bars represent the mean and standard deviations of the treated (blue) and control (red) groups. The dashed lines show linear regression lines for each group. A significant difference in the longitudinal trend associated with treatment was found using a linear mixed effects model (p < 0.001). Figure reproduced with permission from the author and publisher [51].

Discussion

We have summarised the current state of the art with respect to functional imaging methods that have been applied successfully to predict tumour response, duration of local control and overall survival in thermotherapy trials. We have also summarised methods that have potential in this regard, but have not yet been applied in this setting. A significant association has been found between several functional imaging methods and progression-free and overall survival. With the exception of DCE-MRI, however, there is little overlap between which parameters are predictive of which endpoint. Although changes in ATP/Pi appear to reflect direct changes in cell viability, such changes are not related to the ultimate outcome of the patient.

One of the challenges in functional imaging is the complexity and cost of MRI/MRS and PET methods. This becomes particularly problematic when one considers doing repeated measurements on the same subject over time. In our clinical series, the overall success of being able to obtain only two serial measurements was around 50% in both human and canine trials [37], [47]. Measurements made before treatment are relatively straightforward, since there is typically ample time to schedule the measurement before treatment starts. However, once treatment begins, obtaining a follow-up measurement becomes more difficult. Scheduling issues and patient compliance can all contribute to difficulties in obtaining serial imaging data. The rare occurrence of imaging equipment malfunction can also contribute to lost data.

One solution to this problem would be to consider use of optical methods, which are relatively cost effective and rapid to perform. They can be done in the clinic with a portable system at the bedside. Thus, they may provide the most robust tool to obtain serial data, at least for those tumours that are amenable to being studied with this modality.

With regard to molecular markers of outcome, there are important questions regarding whether physiological parameters, such as those discussed in this paper, can be predicted by genomics. Fundamentally, is the physiology of a tumour predetermined by its genomics, or is the physiology independent of this? This relationship has been addressed in multiple myeloma where in a study of over 200 patients the loss of FDG PET signal after induction chemotherapy was a positive predictive factor, and independent of genomic profile [53]. Tzike examined both 1-H MRS parameters and genomic data to classify primary brain tumours [54] and although there were associations between the 1-H MRS parameters and genomics whether these markers were related to outcome was not assessed. To date, there is no information on the relationship of tumour physiology vs. genomics in hyperthermia trials. This is potentially important since a genomic biomarker would be much easier to implement compared with functional imaging.

Expansion of the role of hyperthermia for human cancer treatment will ultimately rely on identification of tumour sites and treatment prescriptions that lead to improved outcome compared to current treatments. Given the time requirements and expense of conducting phase III trials, and the technical effort required to initiate hyperthermia treatment programmes, information that can facilitate identification of patients likely to succeed, or more efficacious treatment prescriptions, could help streamline the definition of optimal hyperthermia treatment settings. Functional imaging has great potential to play an important role in this regard and its potential to modify current hyperthermia treatment practice is only beginning to be tapped.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.