Prediction of Locoregional Control in Head and Neck Squamous Cell Carcinoma with Serial CT Perfusion during Radiotherapy

Discussion

CTP is a functional imaging tool that can supplement conventional CT for mapping the vascularity of the tumor environment. In our study, we found that in patients with HNSCC, blood flow as measured by CTP increased during the 2nd week at 12.72 Gy and then decreased during the 4th week at 33.92 Gy of RT in patients who achieved LRC, whereas patients who had LRF after RT had a decrease in BF relative to their baseline value at 12.72 Gy and then an increase at 33.92 Gy. BF in patients with LRC was noted to rise again at 55.12 Gy and then to subside in the post-RT period compared with that in patients with LRF. These results suggest that a higher BF in tumor tissue at baseline and early during the course of RT resulted in improved tumor control with RT.

Tumors that demonstrated higher baseline BF may represent regions of improved oxygenation and a greater sensitivity to radiation-induced free radical damage. Our study also suggests that in patients with LRC, if BF increases from baseline during the initial course of RT, this change may result in an increased sensitivity of tumor cells to radiation-induced damage with each fraction of RT. In contrast, in patients with LRF, as the BF decreases during RT, this change may further increase the radioresistance of the tumor as patients go through each fraction of RT.

Our study findings for the baseline perfusion values are consistent with the findings of Hermans et al(8) demonstrating, in a study of 105 patients treated with definitive radiation therapy for HNSCC, that the perfusion rate by using dynamic CT before treatment was a predictor for local outcome with lower pretreatment perfusion rates of ≥83.5 mL/min/100 g correlated with decreased local control. Our study findings for the baseline perfusion values for BF of 118.0 mL/min/100 g in patients who were locally controlled were higher than the 83.5 mL/min/100 g cutoff point, compared with patients who were not locally controlled (53.4 mL/min/100 g). Furthermore, our study demonstrated that BF parameters during the course of RT remained higher than the 83.5 mL/min/100 g cutoff value throughout the duration of RT in patients with LRC. Bisdas et al(9) also showed the long-term predictive value of baseline BF and CP measurements from CTP imaging for local control in 84 patients undergoing chemoradiation. Higher BF and CP values significantly correlated with local control compared with LRF. Patients in our cohort who had LRC also had smaller mean tumor volumes compared with patients with LRF. These findings were similar to those in a series of 19 patients with oropharyngeal cancer undergoing induction chemotherapy, in which there were significant correlations between pretreatment tumor volume and baseline BF and MTT perfusion parameters, which predicted progression-free survival.(10) Smaller tumors may have areas of improved BF and oxygenation compared with larger tumors, which often outstrip their blood supply, resulting in a greater hypoxic fraction and decreased perfusion.

While studies also have shown altered perfusion parameters within malignant tissues of the head and neck before undergoing therapy, few studies examine the variations in perfusion parameters during RT. Surlan-Popovic et al(6) studied 20 patients with locally advanced HNSCC undergoing 2 CTP scans after 40 and 70 Gy of chemoradiotherapy. Patients who were classified as responders showed a significant reduction in BF noted after 40 and 70 Gy. Similar perfusion values can be found in our study at 33.92 Gy and 6 weeks after RT (69.96 Gy), when BF was decreased. Our study, however, measured additional time points that demonstrated an initial increase in BF at 12 Gy and then again at 55 Gy. Our results show that BF fluctuates during the course of RT and may be highly dependent on the cumulative therapeutic radiation dose and relative timing of the CTP. This may explain variations of results in these studies, depending on the design of the study and at which time and dose fraction the CTP measurements were made. The initial rise in BF at week 2 followed by a decrease at week 4 and then a rise again at week 6 shows similarities to animal models of depopulation of oxygenated cells with RT and subsequent tumor reoxygenation due to fractionation of radiation therapy.(11) The fluctuations in perfusion parameters in patients with HNSCC at the different dose points may be similar to those in animal models of tumor response to fractionation, demonstrating that the oxygenation of the tumor is constantly changing as oxygenated cells depopulate and hypoxic areas reoxygenate after each radiation dose.(11)

Surlan-Popovic et al(6) additionally showed that a nonsignificant increase in BF, BV, and CP was seen in patients with LRF at 40 and 70 Gy, which was difficult to ascertain from our study due to the limited number of treatment failures and decreasing number of scans at weeks 4 and 6 of treatment.

In the current study, most patients underwent IMRT by using a simultaneously integrated boost or “dose-painted” technique, while in the study by Surlan-Popvic et al,(6) all patients underwent 3D conformal RT. This method of IMRT results in a more heterogeneously delivered radiation-dose distribution to the tumor compared with 3D conformal radiation therapy, which delivers a uniform dose distribution, and this may be a contributing factor to the variation in our CTP results.

The relationship between perfusion characteristics and regions of oxygenation remain unclear, though the hypothesis is that increase in BF, BV, and CP characteristics are a result of upregulation of angiogenic factors such as VEGF in response to hypoxia to promote angiogenesis and in turn improve the oxygenation of the tumor. This finding is supported by studies that have shown a correlation between pretreatment VEGF levels and CTP characteristics, which show a borderline inverse relationship.(12) Studies in vitro of HNSCC cell lines have also shown that RT increases VEGF levels in tumor and endothelial cell lines and that VEGF levels are dependent on cumulative radiation dose, regardless of the intrinsic radiosensitivity of the tumor and baseline VEGF levels.(13)

Variations in perfusion parameters may also be a function of the point of measurement. The current study used 2 tumor measurements obtained in the center of the tumor while avoiding necrotic areas, to minimize variations of the perfusion parameters at the periphery of the tumor that could be caused by inflammation. Studies have demonstrated spatially heterogeneous vascularity within the tumor, and variations between studies may reflect the point of measurement, though most studies attempt to avoid regions of necrotic tissue, large feeding vessels, and normal tissue. In a study of volumetric CTP in patients with non-small cell lung cancer undergoing palliative RT, perfusion measurements at 9, 18, and 27 Gy obtained in 16 patients at the periphery and center of the tumor showed changes in CP and BV at the periphery of the tumor, which were noted to be higher during the course of RT compared with the center. The hypothesis was that the greater changes noted may be due to increased oxygenation at the periphery compared with the center of the tumor, though perfusion measurements at the periphery of the tumor may be difficult to distinguish from the interface with normal tissue.(14)

MR perfusion by using T1-weighted dynamic contrast-enhanced MR imaging shows results similar to those of CTP in a head and neck cancer study of 14 patients, in which an increase in BV was predictive of local control when performed 2 weeks after initiation of RT (equivalent to 20 Gy), whereas BF was not found to be a significant parameter in predicting local control.(15) In this study, however, pretreatment BV and BF did not predict outcome and BV; moreover, with a small patient series and only 9.7 months median follow-up, it is difficult to determine whether these findings correlate with longer follow-up. Other MR perfusion techniques include arterial spin-labeling, which potentially has an advantage over CTP in that magnetically labeled blood water acts as an endogenous tracer obviating contrast administration. This method has been tested in small head and neck cancer series without conclusions regarding the ability to predict outcome but may show promise as a perfusion imaging method for patients who cannot undergo contrast administration.(16)

While concurrent chemoradiotherapy remains the standard of care for locally advanced HNSCC, increasing interest in the role of induction chemotherapy regimens has resulted in re-examination of our current standard treatment paradigms.(17,18) Studies of perfusion imaging before and after induction chemotherapy may further assist patient selection for determining which patients will benefit from concurrent chemoradiation. Zima et al(19) studied 17 patients with HNSCC and found elevated pretreatment BF, BV, and CP by using CTP in patients who achieved >50% response to induction chemotherapy. They used the induction chemotherapy response to then select patients for chemoradiation therapy, and the nonresponders were selected to undergo surgery. Nonresponders were found to have a more tightly clustered distribution of low pretreatment BV levels of 3.0–5.4 mL/100 g, which are similar to the pretreatment BV range found in our LRF group. This study suggests that chemotherapy response may be predicted by pretreatment perfusion parameters, and this may have implications for patient selection for subsequent concurrent chemoradiotherapy.

In contrast, Bisdas et al(20) showed that if patients are selected to undergo surgery followed by adjuvant RT, presurgery perfusion parameters including BF, BF maximum, and CP predict local recurrence. This finding suggests that regardless of treatment, whether it is upfront chemoradiation or surgery followed by adjuvant radiation, baseline tumor perfusion characteristics may predict failure regardless of treatment selection because tumors with unfavorable perfusion characteristics represent a more aggressive phenotype. However, these studies only measured the baseline tumor characteristics at a single time point; and as demonstrated from our study examining tumor perfusion during RT, the perfusion characteristics of the tumor are found to be constantly changing.

As PET has become increasingly integrated into the standard work-up and treatment of HNSCC, integrating CTP could play a role in combination with FDG-PET by correlating tumor perfusion parameters to SUV. PET has the advantage of defined regions of FDG uptake (depending on the threshold used) to allow tumor volume delineation and integration into radiation therapy planning. Small series examining CTP and PET have shown correlations among BF, metabolically active tumor volume, and local control, though FDG uptake by using maximum SUV was not well-correlated with local control in HNSCCs.(21) However, SUV and perfusion characteristics do demonstrate characteristics different from those in normal tissues and larger studies are needed to show the relationship between perfusion parameters and SUV and outcome.(22)

The limitation of our study and of similar studies examining CTP in head and neck RT includes the small patient numbers studied, which precluded further multivariate analysis to adjust for other factors such as staging and volumetric considerations. Other limitations of CTP include the requirement of timed contrast and adequate renal function. In our study, we found that renal insufficiency toward the latter half of chemoradiation therapy related to cisplatin-based therapy, and dehydration may preclude CTP in the last 2 weeks of therapy, as patients experience increased radiation- and chemotherapy-related acute toxicities, making it difficult for the patient to undergo CTP. Increasing concern regarding radiation dose to the normal tissues in patients undergoing CTP has led to using lower kilovoltage settings. Although 80 kV has been used for brain perfusion as well as more recent neck perfusion studies,(23) 120 kV with 60 mA has been used for most published CTP studies for head and neck cancers;(5,8,9,24) therefore, we used 120 kV in this study with 50 mA, decreased from 60 mA. Future studies could be performed with lower kilovoltage and milliampere settings to minimize radiation if appropriate perfusion data could be maintained. Furthermore, MR perfusion is an alternative tool for noninvasive tumor perfusion measurement without ionized radiation exposure, though MR images cannot be obtained in the radiation-treatment position with immobilization devices and MR imaging is contraindicated in patients with cardiac pacemakers and other implanted metallic devices as well as metallic foreign bodies in the brain, globe, and spinal canal.

Future studies could investigate the feasibility of integrating CTP into RT planning by combining the study of anatomic changes during RT with CTP parameters, by improving radiation dose coverage to tumor, or by RT dose escalation to improve LRC. Currently, the CTP datasets are too large to be integrated into current RT planning software. Additionally, CTP is limited in the craniocaudal direction for complete image acquisition of the head and neck region. In our study, the CTP scan encompassed only a 4-cm cranial caudal extent of the tumor; newer CT scanners may allow the entire neck coverage, though this would likely also increase the radiation exposure from the diagnostic scan to the patient. If these technical limitations could be overcome, CTP data could identify areas of increased or decreased tumor perfusion during RT and could be integrated into adaptive radiation therapy such that areas of decreased BF in tumor representing hypoxic regions could potentially receive higher doses with dose-painted IMRT.

For patients who do not achieve LRC with concurrent chemoradiotherapy, pretreatment CTP data demonstrating areas of decreased BF in the tumor may be considered for more aggressive therapy, such as induction chemotherapy followed by concurrent chemoradiotherapy or upfront surgery followed by postoperative radiation therapy. Thus CTP data could have potential implications for improving the selection of patients with HNSCC for chemoradiotherapy.