A retrospective, deformable registration analysis of the impact of PET-CT planning on patterns of failure in stereotactic body radiation therapy for recurrent head and neck cancer

Salvage options are limited for recurrent, previously-irradiated squamous cell carcinoma of the head and neck (rSCCHN), resulting in an overall poor prognosis with a reported median survival of less than one year [1]. For rSCCHN, surgical salvage remains the gold standard but is only possible in approximately 20% of patients [2, 3]. Re-irradiation as a salvage strategy is frequently limited by the normal tissue tolerance of previously-irradiated head and neck structures such as the brainstem, mandible, and spinal cord, but has resulted in a median survival of 9–14 months [4–8]. Stereotactic body radiation therapy (SBRT) has seen increasing use as a salvage strategy for inoperable cancers due to its ability to deliver potent doses of radiation to focal areas with sub-millimeter accuracy, minimizing injury to previously-irradiated tissues and shortening treatment courses. Studies on SBRT for rSCCHN at the University of Pittsburgh Cancer Institute (UPCI) and other institutions have demonstrated control rates comparable to conventional radiotherapy but with fewer toxicities [9–12].

In patients previously treated with surgery, chemotherapy, and/or radiation, delineation of recurrences can be challenging. Treatment planning involves manual delineation of tumors, and can be performed on an integrated PET-CT scan, a PET-aided CT scan, or a CT scan alone. PET-CT has been shown in primary disease to be more sensitive than CT, PET, or PET-aided scans [13]. Whereas in primary SCCHN it is accepted practice to contour regions of lymphatic drainage and sizeable margins around the gross tumor volume (GTV), there is no current standard regarding the use of such margins for recurrent tumors [8]. Current studies on SBRT for rSCCHN, however, have used small or no margins to minimize toxicity [9, 10, 12, 14]. Hence, PET-CT is ideal for delineation in SBRT, whose sharp dose fall-off makes accurate targeting essential. There is yet no data showing that PET-CT is better than non-PET-CT in SBRT treatment planning, judged best by treatment outcome and patterns of failure. We hypothesized that PET-CT planning would lead to a reduction in overall and in particular “near miss” failures, i.e. failures at the PTV border.

The most critical question in any analysis of radiation therapy failures is “Was the tumor missed?” When recurrences develop close to SBRT targets (with rapid dose fall-off outside targets), this question can be very difficult to answer because follow-up scans are usually done with patients in a different position from when they were treated. To overcome this obstacle and analyze patterns of failure, we used the deformable registration function of the VelocityAI^TM [Velocity Medical Systems, Atlanta, GA] software program to compare the location of recurrences on post-treatment scans to the PTV from planning scans. Failures within the PTV may imply resistant tumors or insufficient dose, whereas failures near PTV borders could represent insufficient margins or failure of imaging to detect tumor edges. Few studies exist on patterns of failure after SBRT and the current study is unique in the degree of accuracy we attempt to achieve in using deformable registration to identify the location of recurrences.

Interest in SBRT for rSCCHN has grown due to its ability to deliver high doses of radiation to focal areas with high precision, potentially reducing toxicity and shortening treatment courses [9–12]. This study serves as the first analysis of patterns of failure after SBRT for rSCCHN, and thus provides crucial information for advancing the field and enhancing efficacy of this novel targeted therapy. The availability of many bony landmarks in the head and neck enhances accuracy of deformable registration and allows a detailed pattern of failure analysis. Our failure categories were, in order of proximity to the PTV: In-field, Overlap, Marginal, and Regional/Distant. We found that most failures were on or near PTV borders (Overlap/Marginal failure, 61.4%), rather than within the PTV (In-field failure, 12.3%) or over 1 cm from the PTV (Regional/Distant failure, 26.3%).

Whereas most prior pattern of failure analyses of radiation for primary head and neck SCC showed mostly in-field recurrences [21, 22], we found a high frequency of edge-of-field failures after SBRT for rSCCHN. Our findings also stand in contrast to one of the few other existing studies on patterns of failure after re-irradiation of rSCCHN, which used IMRT and 3-D conformal radiation therapy rather than SBRT. In that study, Popovtzer et al. reported a large majority of in-field recurrences and suggested that confining radiation targets to the GTV could reduce toxicity without compromising overall control [23]. However, that study used a 5 mm margin around the GTV (vs. no margin in our study) and a less stringent definition of in-field recurrence (>50% of recurrence within 95% isodose curve vs. >75% of recurrence within PTV in our study), possibly resulting in a lower frequency of near miss failures. When we applied Popovtzer et al.’s definitions to our database, Overlap/Marginal failures still accounted for 53% of failures.

Treatment of recurrent head and neck cancer differs from primary tumors in that for recurrences, prophylactic contouring of regions of lymphatic drainage and large margins are not standard.⁸ With the smaller targets and the sharp dose fall-off involved in SBRT for rSCCHN, the potential for near misses increases and accurate detection and contouring of tumor edges becomes critical. Detection of tumor edges may be enhanced with PET-CT treatment planning. We found PET-CT planning to be associated with longer overall failure-free survival and longer Overlap/Marginal failure-free survival after SBRT for rSCCHN, suggesting an advantage in tumor control. This association was stronger in the definitive SBRT subgroup, which likely reflects the higher usage of CT planning in adjuvant treatments for positive surgical margins included in the overall cohort. For time to combined In-field/Overlap/Marginal failure, PET-CT showed no significant benefit in the overall cohort and only a borderline benefit in the definitive SBRT subgroup. This suggests that PET-CT planning improves control principally through reduction of failures on the PTV edge.

Overall survival for patients without distant metastases or untreated locoregional disease in this study was similar to prior reports on SBRT for rSCCHN [9–12]. We found that older age, lower performance status (KPS), presence of distant metastases/untreated locoregional disease, higher tumor volume, and lack of prior surgery were significant predictors of decreased overall survival. We did not observe a significant difference in overall survival between the PET-CT and non-PET-CT planning groups.

A primary limitation of the current study is our inability to detect precisely where recurrent tumors originated. Recurrent tumors are often large by the time they are detected and if any part of the recurrence overlaps with the PTV border, clinical judgment becomes important in defining whether the recurrence originated from inside or outside the PTV. This may lead to subjective categorization of failures as in-field vs. out of field and subsequent variation in definitions of category of failure. Other limitations include patient heterogeneity and potential variability in contouring recurrent tumors.

This study demonstrates that most recurrences after SBRT for rSCCHN were near misses, i.e. failures occurring near the PTV border. PET-CT treatment planning was shown to result in lower rates of failure, particularly near miss failures, compared to non-PET-CT planning. PET-CT appears advantageous for treatment planning in SBRT for rSCCHN, where accurate targeting of smaller tumor volumes within previously-irradiated tissues is critical.