|Home | About | Journals | Submit | Contact Us | Français|
For the treatment of squamous cell cancer of the head and neck (SCCHN), the assessment of treatment response is traditionally accomplished by volumetric measurements and has been suggested to be prognostic for an eventual response to treatment. An early evaluation response during the course of radiation therapy could provide an opportunity to tailor treatment to individual patients. Diffusion magnetic resonance imaging (MRI) allows for the quantification of tissue water diffusion values, thus treatment-induced loss of tumor cells will result in the increase in water mobility at the microscopic level, which can be detected as an increase in tumor diffusion values before any volumetric changes occur. We evaluated the use of diffusion MRI as an imaging biomarker of treatment response in an orthotopic mouse model of SCCHN. Mice with murine squamous cells expressing the yeast transgene cytosine deaminase were treated with 5-fluorocytosine (5FC), ionizing radiation, and combined therapy and were compared with control animals both during and after treatment for changes in tumor volumes, diffusion values, and survival. Radiation therapy had minimal effect on volumetric growth rate, diffusion, or survival. Although 5FC and combination treatment resulted in similar reductions in tumor volumes, the combination treatment elicited a much greater increase in tumor diffusion values, which correlated with improved survival. Thus, diffusion MRI as an imaging biomarker has a potential for early evaluation of the response to chemoradiation treatment in SCCHN.
There will be an estimated 47,000 cancers of the mouth, tongue, pharynx, and larynx in the United States in 2008 with approximately 11,000 deaths . As such, squamous cell cancer of the head and neck (SCCHN) will makeup 2.5% of all cancers and 1% of all deaths caused by cancer during this period. Greater than 90% of these cases are composed of the common histologic diagnosis, squamous cell carcinoma. Surgery remains the mainstay of treatment in head and neck cancer; however, more than 60% of all SCCHN are present in a locally advanced stage where surgery alone is no longer considered curative. Therefore, the combination of chemotherapy and radiation therapy either adjuvantly after maximal surgical resection or as definitive treatment has become the standard of care for many patients .
Chemotherapy alone has an excellent response rate in SCCHN with up to 80% overall response and complete response (CR) rates in the range of 25% when using multiagent induction chemotherapy . However, despite these excellent response rates, there is little evidence for a survival benefit with induction chemotherapy followed by either surgery or definitive radiation therapy, although a recent meta-analysis did reveal a small survival benefit to induction chemotherapy before radiotherapy . In contrast, the addition of concurrent chemotherapy to radiation therapy (most typically platinum based chemotherapy) has proven to provide a significant benefit in terms of local control and overall survival in both definitive and adjuvant radiation therapies in the treatment of locally advanced SCCHN [2–4].
Typically, evaluation of response to therapy in head and neck cancer has been through physical examination, evaluation in the operating suite through direct laryngoscopy, and by cross-sectional imaging [computed tomography (CT) or magnetic resonance imaging (MRI)] . An early evaluation of tumor response may provide prognostic information about treatment efficacy and allow subsequent tailoring of interventions based upon individual response such as using imageguided adaptive radiotherapy . Computed tomography has been evaluated as a noninvasive means for evaluating the response of induction chemoradiotherapy before surgical resection for SCCHN. However, in this case, strictly volumetric response based on CT examinations gave both poor positive and negative predictive values when compared to histologic finding obtained at resection . Computed tomographic examinations have also been used to evaluate response to radiation therapy both during treatment and 8 to 10 weeks after the completion of treatment [7–11]. Some authors have suggested that rapid response during radiation therapy was a positive predictor of eventual tumor control , whereas others have found that a more protracted response was a better prognostic factor . Thus, the use of volumetric-based measurements especially during radiation treatment is unclear at this time, although the combined use of CT and endoscopic evaluation has been suggested to predict for a good response to chemotherapy and a greater likelihood of organ preservation in head and neck cancer of the larynx, hypopharynx, and oropharynx [12–16].
Functional imaging.based assessments have also been assessed recently. Positron emission tomography using 18F-fluoro-deoxy-glucose (FDG-PET) has been proven to predict for response to chemoradiotherapy in patients with head and neck cancer when measured several months after treatment [17,18]. In addition, a negative FDGPET scan has been demonstrated to predict response in cervical lymph node metastases of head and neck cancer, therefore eliminating the need for post treatment neck dissection in patients who achieve a CR by FDG-PET . However, currently, there is limited evidence for the use of PET during radiotherapy, and indeed, there is concern that acute inflammation caused by radiation of mucosal tissues could lead to false-positive determinations using FDG-PET . Dynamic contrast-enhanced MRI  or other perfusion-based techniques  are alternative imaging modalities, which have been evaluated as early markers of response in head and neck cancer.
An imaging approach that is rapidly gaining interest for the detection of cancer treatment response is diffusion MRI  (Figure 1). Diffusion MRI is sensitive to the subcellular motion of water in biologic tissues and has been demonstrated to be effective both in diagnosing head and neck cancers  and in detecting relapses many months after the completion of radiation therapy . However, the use of diffusion MRI during treatment of head and neck cancer has not been appropriately evaluated. Diffusion MRI early in a course of treatment would allow for the detection of alterations in tumor cell membrane integrity (i.e., microenvironment) after a therapeutic intervention using the Brownian motion of water (e.g., water diffusion) as a surrogate. This is accomplished based on the use of diffusion MRI pulse sequences allowing for collection of multislice image data sets wherein the magnetic resonance signal intensity is dependent on the mobility of the water molecules within the tissue . At the start of treatment, tumors are anticipated to be densely cellular with a diffusion histogram reflective of this (Figure 1, top panel). Shortly after initiation of an effective treatment, there is initial breakdown of cellular membranes, which is detectable as a shift in the diffusion histogram to the right to more closely approximate an environment with increased water mobility, and this increase in water diffusion precedes overt volumetric changes in tumor size (Figure 1, middle panel). Finally, after the completion of an effective treatment, frank tumor death (necrotic or apoptotic) leads to further breakdown of cells and broadening of the diffusion histogram and often with a coincident decrease in tumor volume (Figure 1, bottom panel).
The initial application of diffusion MRI for the detection of tumor treatment response was reported using a rodent glioma model . Further validation studies have expanded significantly on these initial results using a variety of tumor models and therapeutic agents [27–31]. Overall, these studies have shown that diffusion MRI can serve as a sensitive imaging biomarker for the detection of early cellular changes in treated tumors, which precede macroscopic volumetric response. Moreover, diffusion MRI measurements can be used to identify spatially distinct regional responses to therapy in tumors [31–35]. Translation of diffusion MRI into brain tumor clinical trials has been accomplished [33,34,36] where changes in diffusion measures have recently been shown to be strongly correlated both with overall clinical response based on the World Health Organization response criteria [34,36] and with overall survival . In addition, preliminary data would also support diffusion MRI as an early response metric to combined chemoradiotherapy in head and neck cancer [37,38]. Thus, diffusion MRI is emerging as an important predictive biomarker for early stratification of tumor response.
The purpose of the current study was to evaluate whether diffusion MRI could be extended to assess treatment effects in SCCHN. The murine SCCVII squamous tumor model was used with cells genetically engineered to express a therapeutic transgene [yeast cytosine deaminase (yCD)], which, when exposed to a prodrug, 5-fluorocytosine (5FC), generates high concentrations of the chemotherapeutic agent 5-fluorouracil (5FU) locally within the tumor . Treatment effects on diffusion were evaluated after treatment with radiation and radiochemotherapy and revealed that SCCVII tumor diffusion changes were correlated with treatment outcome measures (tumor burden and survival) better than conventional radiographic response, suggesting that further translational clinical studies are warranted.
All of the animal work was carried out in the animal facility at the University of Michigan in accordance with federal, local, and institutional guidelines. Briefly, a yCD-expressing murine squamous SCCVII stable cell line was established by retroviral infection and limiting dilution in 96-well plates as previously reported . These SCCVII cells (grown in medium culture) were trypsinized, washed in phosphate-buffered saline (PBS), counted, and resuspended at a concentration of 4 x 108 cells/ml in PBS. Fifty microliters of this cell solution was injected into the submental compartment of C3H mice (6–8 weeks; Charles River Laboratories, Wilmington, MA) along the midline, using an external approach [39–41]. Mice were weighed every 2 to 3 days and were euthanized if they exceeded a body weight loss of >25%. For all studies, animals were anesthetized with an air/isoflurane (1.25%) mixture.
Twenty-four animals with SCCVII tumors were entered into the study. Treatment was delivered as previously described . When in vivo tumor volumes reached 15 to 20 µl, animals were divided into four groups. Group 1 was designated as the control group that received 0.1 ml of drug vehicle (10% ethanol in PBS daily by i.p. injection; n = 6). Groups 2, 3, and 4 received 5FC (n = 6), radiation (n = 6), and 5FC plus radiation (n = 6), respectively. All 5FC treatments (500 mg/kg daily) were administered by a single i.p. injection on days 0 to 4. Groups 3 and 4 animals received five doses of ionizing radiation at 4 Gy/day delivered on days 0 to 4. T2-weighted and diffusion MRI was performed every other day (n = 3 per imaging group) to measure volumetric and tumor water diffusion changes, respectively. Animal survival data were also obtained for all animals.
Tumors were imaged over a 3 week time frame using a Varian Unity Inova MRI system (Varian Instruments, Fremont, CA) equipped with a 7-T, 18.3-cm horizontal bore magnet (Oxford Instruments, Oxford, UK). During all MRI procedures, animals were anesthetized using 1.25% isoflurane, and body temperature was maintained at 37°C using a heated water recirculating pad. For anatomical imaging, a “fast spin-echo multislice” MRI sequence (TR/TE = 4000/60 milliseconds, number of transitients = 4, number of echoes = 8, FOV = 30 x 30 mm, image matrix = 128 x 128, slice thickness = 1 mm) was used to collect 13 contiguous T2-weighted images.
Maps of tumor apparent diffusion coefficient (ADC) values were acquired using a recently described method . Briefly, a trace diffusion-weighted multislice spin-echo sequence (with motion compensation and a navigator echo) was used to acquire 13 slices with two different diffusion weightings (b1 = 100 and b2 = 1248 sec/mm2). The image slice thickness was 1 mm, image matrix was 128 x 128 (zero filled to 256), field of view was 30 x 30 mm, and echo time was 60 milliseconds. The images acquired with b1 were essentially T2-weighted images, and these were used to segment the tumor from nontumor tissue for volumetric analysis using an “in-house” region drawing tool developed in Matlab.
A 2-tailed t test was used to compare volume and diffusion data between groups. For comparison of animal overall survival, a log rank test was performed. Statistical analysis was performed using MedCalc (MedCalc, Version 9.5.1, Mariakerke, Belgium) with P < .05 considered to be statistically significant.
In this study, we used an orthotopic model of SCCHN, which was previously demonstrated to mimic the locally invasive nature of this disease [39–41]. Murine squamous cells (SCCVII) were injected submentally forming a locally infiltrative tumor with near 100% tumor take. Tumors were allowed to grow sufficiently to have measurable dimensions for MRI-based imaging (15–20 µl at 3 days after implantation; Figure 2). As previously reported, we found that radiation therapy alone (4 Gy x 5 daily fractions) had limited efficacy in this model given the rapid growth of the tumor and the constraints on radiation dose caused by adjacent normal mucosa . Combined chemoradiotherapy in this model using conventional agents (cisplatin, 5FU, or gemcitabine) was also found to be problematic owing to the abrupt and abundant mucositis and corresponding weight loss that developed. We initially tested the above three chemoradiotherapy regimens in non-tumor-bearing mice with radiation administered to the oral mucosa, and median weight loss within the first 2 weeks was >25% with a significant number of animals requiring euthanasia secondary to treatment toxicity (data not shown). We thus engendered to find an alternative model for concurrent chemoradiotherapy, which would limit the acute mucosal reactions and morbidity and allow evaluation of early MRI markers of treatment response [39,41].
Previously, we have reported on the use of genetically modified cells stably expressing either bacterial  or yeast cytosine deaminase , which allow the local conversion of 5FC, a nontoxic prodrug, to 5FU, a commonly used chemotherapeutic and radiosensitizing agent. Use of this gene therapy strategy was tolerable in this orthotopic model either alone or when combined with fractionated external beam radiotherapy [39,41]. Therefore, to serve as a model for concurrent chemoradiotherapy, orthotopic tumors that expressed yeast cytosine deaminase were developed, which were then treated with systemic 5FC either with or without fractionated external beam radiotherapy. After implantation, tumors expressing yCD readily grew within the submental compartment with a tumor-doubling time (4.1 ± 0.8 days) that was indistinguishable from wild type SCCVII cells [40,41] (P > .1; Figures 2A and and33).
To test the use of radiation therapy either alone or when combined with systemically administered 5FC in yCD-expressing tumors in the submental compartment, animals were separated into four treatment groups. Control animals received a daily i.p. injection of carrier solution for 5 days. Radiation-treated animals were treated with 4 Gy daily for five consecutive days. Chemotherapy-treated animals received five consecutive daily i.p. injections of 5FC (500 mg/kg). Finally, combined modality-treated animals received both daily 5FC for 5 days along with five fractions of radiation (4 Gy each) delivered concurrent with chemotherapy treatment. A total of six animals were treated in each group for toxicity and survival analysis, and at the start of the study, three animals from each group were randomly selected for sequential imaging with MRI.
Figure 2 displays sequential representative T2-weighted MRI slices from one animal from each group, whereas Figure 3 displays the overall tumor growth curve. The maximum tumor diameter is depicted at each time point on the images, although all measures of response were based on volumetric analysis across all MRI slices. These diffusion-weighted images are essentially heavily T2-weighted with areas of increasing diffusion depicted as a higher intensity signal. As can be seen in Figure 2 A, control animals had tumors that grew rapidly after implantation with a locally invasive and exophytic tumor developing within the submental compartment. These tumors were also relatively solid with an ADC, which approximated that of surrounding muscle (Figure 2A); however, during the first week, as untreated tumors grew (Figure 3), they become more solid and cellular with a decrease in the ADC (Figure 4). Treatment with fractionated radiation therapy alone had only limited impact on tumor growth (Figure 3) that was not significantly different from control animals. In addition, the diffusion characteristics of these tumors approximated surrounding soft tissues with no significant increase in diffusion signal (Figure 4). In animals treated with systemic 5FC (Figure 2B), there was a marked arrest of tumor growth within 3 days after treatment, with a complete regression of tumor witnessed by day 7 of treatment; however, there was no significant increase in diffusion signal within the tumors during this period of tumor regression (Figure 2B). Finally, in animals treated with combination therapy (Figure 2C), there was also a significant regression of tumor leading to a CR by day 7 after treatment, which paralleled the 5FC-treated animals in the rate and extent of response (Figure 3). However, unlike the 5FC-treated animals, there was a substantial rise in the ADC of the submental tumors as witnessed by the greatly increased signal within the tumor (Figure 2C).
A comparison of the rate of tumor growth across all animals revealed a near constant growth of control tumors reaching a volume more than threefold larger than the starting size 9 days after the start of treatment (and 12 days after implantation; Figure 3, open squares). In animals treated with daily fractionated radiation, there was only limited growth delay that was not statistically significant when compared with control tumors (Figure 3, closed squares). In contrast, treatment with daily 5FC (Figure 3, open circles) resulted in rapid tumor shrinkage such that no measurable tumor was identifiable as early as 9 days after the start of treatment. Finally, in animals treated with combination therapy (Figure 3, closed circles), there was also an abrupt shrinkage of tumor resulting in a radiographic CR by 9 days after treatment. The rates of tumor regression in 5FC- or combinationtreated animals were both statistically different from the control- and radiation-treated groups from day 3 of treatment onward (all P values in pairwise comparison <.04). In contrast, there was no significant difference between the growth rates of control- and radiation-treated animals (P > .1) or between the 5FC or combination therapy groups (P > .1).
Unlike the volumetric measures, which could not distinguish between the two treatment groups involving chemotherapy, when the mean ADC in the tumor masses were calculated, it was apparent that the combination therapy animals were exhibiting a much more abrupt rise in ADC (Figure 4, closed circles), which was different from the other three groups by 5 days after treatment and this difference continued to increase. The area under the curve of the change in the ADC as a function of time for combination therapy was statistically different from all other treatment groups (P < .03, P < .04, and P < .05, respectively, when compared with control-, radiation-, and 5FC-treated groups). In contrast, there was no significant difference between the remaining treatment groups; although there was a trend toward a greater area under the curve for 5FC-treated animals when compared with radiation-treated or control animals (P < .08).
Overall tumor control and animal survival were evaluated, and there was no significant difference in tumor control or median survival between control- and radiation-treated groups (Figure 5). There was a trend toward a slightly increased median survival time in animals treated with radiation when compared with controls (33 vs 28 days); however, this difference did not achieve statistical significance (P > .1). In contrast, in animals treated with daily 5FC, there was improved tumor control, with two of six animals free of tumor 90 days after tumor implantation and a median survival of 66 days that was statistically different from both control- and radiation-treated animals (P < .0006 for each). Finally, in the combination therapy group of animals, a greatly enhanced tumor control was observed when compared to all other groups, with five of six animals free of tumor 90 days after implantation and the median survival time that was not reached. Survival in the combination treatment group was statistically improved when compared not only to control- or radiation-treated animals (P < .0006 for each) but also to those animals treated with 5FC only (P < .05). At the time of death of all animals in each of these four groups, there was a grossly identifiable tumor within the submental compartment at postmortem analysis. In contrast, at the end of the experiment (90 days after implantation), there was no radiographically or grossly identifiable disease in any of the seven animals left alive in the 5FC or combination therapy groups.
Treatment of head and neck cancer once the province of surgical resection alone is becoming more and more a multimodality disease [2,3]. However, the best way to integrate these different modalities is still not completely clear. Each of the commonly used treatments (surgery, radiation therapy, and chemotherapy) has its own unique benefits and toxicities. The goal of multimodality therapy is to judiciously use each of these treatments in a manner that maximizes benefit while limiting toxicity. The goal of early assessments of treatment response would be to allow an early and accurate evaluation of treatment efficacy, which would enable treatments to be tailored between individual patients instead of simply applying a treatment based on population norms or historical treatment information.
Standard volumetric-based assessments of treatment response in patients with head and neck cancer after induction chemotherapy have enabled selection of patients who are likely to benefit from subsequent radiotherapy or chemoradiotherapy [12–14,16]. However, these regimens have often entailed as many as three cycles of chemotherapy delivered for 9 or more weeks, which greatly increased the time and morbidity of treatment without providing any improvement in patient survival [12–14]. The use of volumetric measures during or shortly after radiotherapy has been difficult and has been found to only poorly correlate with eventual pathologic tumor control. In contrast, response as assessed by CTscan or PETmany weeks to months after radiotherapy does have prognostic information regarding tumor control but is hampered by the long time after treatment to assess the response and would not allow adaptive radiotherapy based on intratreatment assessment of response.
In the current study, we tested a novel imaging technique to compare the differences in diffusion-weighted imaging within a tumor mass in response to treatment as a metric for eventual tumor control. The model system used incorporated a rapidly growing SCCHN line that is only minimally responsive to radiotherapy as revealed by both standard MRI  and diffusion MRI. Either chemotherapy, using a novel enzyme/prodrug gene therapy system, or the combination of chemotherapy and radiotherapy gave excellent response with complete regression of tumor observed in all animals. Despite complete radiographic response to single-agent chemotherapy, four of six animals eventually recurred with large submental tumors. Supporting our earlier work with this system [39,41], the combination of radiotherapy and chemotherapy did result in increased tumor control and enhanced animal survival compared with either chemotherapy or radiotherapy alone with five of six animals alive and free of disease 90 days after tumor implantation.
Diffusion MRI has previously been demonstrated to correlate with cellular density in rodent models of glioma and breast cancer [42–44]. In addition, changes in diffusion-weighted imaging have been proposed as a marker for early treatment response in both animal models of glioma [35,43] and in patients receiving chemotherapy or radiotherapy for primary brain tumors [33,34,36,45,46]. In patients with primary central nervous system malignancies, the combination of diffusion-weighted imaging both before and 3 weeks after the start of treatment has been demonstrated to predict eventual radiographic response at 10 weeks with both high sensitivity and specificity . In addition, in a subset of patients with high-grade malignant glioma, diffusion-weighted imaging obtained as early as 3 weeks after the start of a course of fractionated radiation therapy predicted not only radiographic response but also time to progression and overall survival [33,34]. In our studies of glioma, we have found that a regional evaluation of diffusion response was more predictive of survival than simple measurements in the increase in overall average ADC [33–36]; however, in this animal model, given the very small size of tumors and their rapid growth rate, this regional evaluation was not feasible. Therefore, it is possible that, with further technologic application, this regional evaluation of diffusion response will be even more predictive of later clinical outcome.
One further potential advantage demonstrated in this study, which needs to be explored, is the fact that, even with CR to therapy, 5 of 12 animals eventually recurred (4/6 in the 5FC group and 1/6 in the 5FC + radiation group). In contrast, the much greater rise in ADC observed in the combined modality group only 3 days into treatment correlated with a significantly greater tumor control. We previously observed this phenomenon in a rodent glioma tumor model as well where radiographic CR to treatment was poorly correlated with tumor control, whereas a radiographic CR in the setting of a threefold rise in tumor ADC predicted for long-term tumor control . The lack of a CR to chemotherapy correlating with significant clinical advantage would also be supported by the extensive experience with induction chemotherapy in head and neck cancer where as many as 30% to 40% of patients could be rendered radiographically free of disease before radiotherapy with only a very modest improvement in survival . In contrast, a more intensive therapy with concurrent chemotherapy and radiation therapy has been associated with a significantly greater survival advantage .
Thus, in preclinical and clinical evaluation of head and neck cancer, a rapid CR to chemotherapy did not predict for ultimate tumor control. However, the preclinical data presented here for diffusion MRI suggest that this technique may be able to differentiate between those radiographic responses that are more likely to be durable. Initial experience with diffusion MRI in head and neck cancer supports this technique as technically achievable in patients with head and neck cancer being treated with concurrent chemoradiotherapy. In two pilot studies, significant increases in tumor ADC were observed as early as 1  to 2 weeks  into chemoradiotherapy that seem to predict subsequent radiographic response. It remains to be seen if further evaluation of diffusion MRI as an early biomarker of response will enable patient-specific tailoring of chemotherapy and radiation therapy in patients with head and neck cancer to maintain tumor control while decreasing the overall toxicity of treatment and increasing patient quality of life.
1D.A.H. was supported by an ASCO Foundation Young Investigator Award. Support was also provided by a National Institutes of Health grant (P50CA01014).