2.1

Patient selection

This retrospective study was approved by our institutional review board. A total of 34 patients treated with definitive CRT between 2008 and 2014 for Stage III–IV HNSCC using IMRT on our tomotherapy linear accelerator (Madison, WI) were retrospectively reviewed. Patients were included only if they had radiographically identifiable disease at the start of therapy, defined as having measurable PT and/or ND measuring ≥ 2.5 cm ³ and ≥ 1 cm in short-axis diameter. We excluded patients if they had received neoadjuvant chemotherapy, had non-squamous cell histology, were diagnosed with multiple synchronous head and neck primary cancers, or suffered treatment breaks that lasted longer than two weeks during their therapy.

2.2

IMRT technique

All 34 patients underwent diagnostic PET/CT scan prior to treatment. CT simulation images were acquired with a Philips Brilliance Big Bore scanner (Andover, MA) in 3 mm contiguous slices from skull vertex to mid-thorax after 100 cm ³ of isoview contrast (Bracco Imaging, Monroe Township, NJ) was administered. IMRT planning was performed with simultaneous integrated boost utilizing 2–3 dose levels for each patient. Macroscopic gross disease with additional margin accounting for microscopic disease, internal motion, and daily set-up uncertainties (planned target volume [PTV 1]) was contoured and prescribed to 70 Gy in 33–35 fractions. High risk microscopic regions and nodal levels (PTV2) were prescribed a dose ranging from 59 to 63 Gy in 33–35 fractions. Low risk elective nodal regions (PTV3) were prescribed a dose ranging from 52 to 56 Gy in 33–35 fractions. All PTVs were prescribed to a minimum of 95% coverage. Thirty three patients underwent post-treatment PET/CT scan at a median time of 101 days (range: 62–444 days).

2.3

Volumetric analysis

All patients underwent daily image guidance with MVCT prior to each fraction of treatment. MVCT images were acquired utilizing the gantry mounted system with an energy of 1 MeV, dose rate of 266–283 cGy per min, slice thickness of 5 mm, and field-of-view extending the length of PTVs contoured. The estimated additional dose delivered to the patient ranged from 1 to 4 cGy per daily MVCT. A total of 7 weekly representative MVCT image sets were extracted per patient and were imported from tomotherapy dataserver into Pinnacle planning software (Philips Healthcare, Andover, MA). Week 0 MVCT represented images taken at the start of therapy and week 6 represented those taken at the end of therapy. Delineation of identifiable PT and/or ND was performed on each patient’s image sets utilizing head and neck windowing. When contouring on week 0 MVCT imaging, disease was delineated with the assistance of pre-treatment hypermetabolic findings and enhancing portions from diagnostic PET/CTs and with contouring fused from the original treatment plan. For subsequent serial MVCTs, the window leveling which was utilized for week 0 for a given patient was fixed in effort to reduce interpreting variability. Given the poorer soft tissue contrast with MVCT versus kilovoltage CT, changes in more obvious asymmetric and exophytic components of disease served as a surrogate for contouring changes in contiguous but less obvious infiltrative and endophytic portions of disease.

PT was contoured only when an obvious asymmetric or exophytic component of ≥ 2.5 cm ³ and ≥ 1 cm in short axis was observable on week 0 MVCT. ND was contoured only when an asymmetric component of ≥ 2.5 cm ³ and ≥ 1 cm in short axis was seen when compared to the contralateral neck at that level. When the patient presented with multiple pathologic nodal involvement, only the largest node or nodal conglomerate was chosen for MVCT contouring. In 19 of 34 total patients, both the PT and ND volumes were identifiable and therefore contoured. In 6 patients, only the PT was contoured because nodes were not clearly visible on MVCT ( n = 1) or the patient had node-negative disease ( n = 5). In 9 patients, only the ND was contoured since the PT was not clearly visible on the MVCT. Contouring for each patient was performed while blinded to patient outcome to minimize bias. Given time constraints, a resident physician (RD) contoured in tandem with the head and neck radiation oncologist (SY) for averaging effects of potential interobserver variability.

2.4

Outcomes and variables analyzed

Patient characteristics that were analyzed included smoking status, p16 status, T stage, N stage, and whether the patient required a break from therapy. Disease related outcomes analyzed included percent residual disease volume, rate of disease regression, local failure (LF), progression-free survival (PFS), and overall survival (OS). LF was defined as persistent or recurrent disease within the specified contoured volume on serial MVCTs. Treatment failures were confirmed either as clinically and radiographically enlarging masses or by pathologic assessment.

2.5

Statistical analysis

Because of large initial disease volume differences between patients, PT and ND volumes on serial MVCTs were normalized to week 0 volumes per patient to evaluate relative volume changes. PT ( n = 25) and ND ( n = 28) data were then analyzed separately when calculating percent residual volume at week 6 and regression rates. Differences in percent residual volume at week 6 between those with and without LF were compared using student t -test. Regression rates were estimated using linear regression modeling, and the differences in regression rates between patients with and without LF were compared using t -test.

An optimal cut-off value of 25% residual volumes for both PT and ND was then established by the Youden index method. The relative risks (RR) for LF and their 95% confidence intervals utilizing the 25% residual threshold were estimated and tested using Fisher exact test. Positive predictive value (PPV), negative predictive value (NPV), sensitivity, specificity, and accuracy for predicting LF were calculated while utilizing the 25% residual volume cut-off value.

When analyzing PFS and OS, patients were stratified according to whether they had greater or less than 25% residual volume in either the PT or ND at the end of treatment. For this portion of the analysis all patients were evaluated in aggregate and Kaplan–Meier analysis was used to estimate the PFS and OS. Log-rank test was then employed to test the differences in the survival curves.

2.1

Patient selection

This retrospective study was approved by our institutional review board. A total of 34 patients treated with definitive CRT between 2008 and 2014 for Stage III–IV HNSCC using IMRT on our tomotherapy linear accelerator (Madison, WI) were retrospectively reviewed. Patients were included only if they had radiographically identifiable disease at the start of therapy, defined as having measurable PT and/or ND measuring ≥ 2.5 cm ³ and ≥ 1 cm in short-axis diameter. We excluded patients if they had received neoadjuvant chemotherapy, had non-squamous cell histology, were diagnosed with multiple synchronous head and neck primary cancers, or suffered treatment breaks that lasted longer than two weeks during their therapy.

2.2

IMRT technique

All 34 patients underwent diagnostic PET/CT scan prior to treatment. CT simulation images were acquired with a Philips Brilliance Big Bore scanner (Andover, MA) in 3 mm contiguous slices from skull vertex to mid-thorax after 100 cm ³ of isoview contrast (Bracco Imaging, Monroe Township, NJ) was administered. IMRT planning was performed with simultaneous integrated boost utilizing 2–3 dose levels for each patient. Macroscopic gross disease with additional margin accounting for microscopic disease, internal motion, and daily set-up uncertainties (planned target volume [PTV 1]) was contoured and prescribed to 70 Gy in 33–35 fractions. High risk microscopic regions and nodal levels (PTV2) were prescribed a dose ranging from 59 to 63 Gy in 33–35 fractions. Low risk elective nodal regions (PTV3) were prescribed a dose ranging from 52 to 56 Gy in 33–35 fractions. All PTVs were prescribed to a minimum of 95% coverage. Thirty three patients underwent post-treatment PET/CT scan at a median time of 101 days (range: 62–444 days).

2.3

Volumetric analysis

All patients underwent daily image guidance with MVCT prior to each fraction of treatment. MVCT images were acquired utilizing the gantry mounted system with an energy of 1 MeV, dose rate of 266–283 cGy per min, slice thickness of 5 mm, and field-of-view extending the length of PTVs contoured. The estimated additional dose delivered to the patient ranged from 1 to 4 cGy per daily MVCT. A total of 7 weekly representative MVCT image sets were extracted per patient and were imported from tomotherapy dataserver into Pinnacle planning software (Philips Healthcare, Andover, MA). Week 0 MVCT represented images taken at the start of therapy and week 6 represented those taken at the end of therapy. Delineation of identifiable PT and/or ND was performed on each patient’s image sets utilizing head and neck windowing. When contouring on week 0 MVCT imaging, disease was delineated with the assistance of pre-treatment hypermetabolic findings and enhancing portions from diagnostic PET/CTs and with contouring fused from the original treatment plan. For subsequent serial MVCTs, the window leveling which was utilized for week 0 for a given patient was fixed in effort to reduce interpreting variability. Given the poorer soft tissue contrast with MVCT versus kilovoltage CT, changes in more obvious asymmetric and exophytic components of disease served as a surrogate for contouring changes in contiguous but less obvious infiltrative and endophytic portions of disease.

PT was contoured only when an obvious asymmetric or exophytic component of ≥ 2.5 cm ³ and ≥ 1 cm in short axis was observable on week 0 MVCT. ND was contoured only when an asymmetric component of ≥ 2.5 cm ³ and ≥ 1 cm in short axis was seen when compared to the contralateral neck at that level. When the patient presented with multiple pathologic nodal involvement, only the largest node or nodal conglomerate was chosen for MVCT contouring. In 19 of 34 total patients, both the PT and ND volumes were identifiable and therefore contoured. In 6 patients, only the PT was contoured because nodes were not clearly visible on MVCT ( n = 1) or the patient had node-negative disease ( n = 5). In 9 patients, only the ND was contoured since the PT was not clearly visible on the MVCT. Contouring for each patient was performed while blinded to patient outcome to minimize bias. Given time constraints, a resident physician (RD) contoured in tandem with the head and neck radiation oncologist (SY) for averaging effects of potential interobserver variability.

2.4

Outcomes and variables analyzed

Patient characteristics that were analyzed included smoking status, p16 status, T stage, N stage, and whether the patient required a break from therapy. Disease related outcomes analyzed included percent residual disease volume, rate of disease regression, local failure (LF), progression-free survival (PFS), and overall survival (OS). LF was defined as persistent or recurrent disease within the specified contoured volume on serial MVCTs. Treatment failures were confirmed either as clinically and radiographically enlarging masses or by pathologic assessment.

2.5

Statistical analysis

Because of large initial disease volume differences between patients, PT and ND volumes on serial MVCTs were normalized to week 0 volumes per patient to evaluate relative volume changes. PT ( n = 25) and ND ( n = 28) data were then analyzed separately when calculating percent residual volume at week 6 and regression rates. Differences in percent residual volume at week 6 between those with and without LF were compared using student t -test. Regression rates were estimated using linear regression modeling, and the differences in regression rates between patients with and without LF were compared using t -test.

An optimal cut-off value of 25% residual volumes for both PT and ND was then established by the Youden index method. The relative risks (RR) for LF and their 95% confidence intervals utilizing the 25% residual threshold were estimated and tested using Fisher exact test. Positive predictive value (PPV), negative predictive value (NPV), sensitivity, specificity, and accuracy for predicting LF were calculated while utilizing the 25% residual volume cut-off value.

When analyzing PFS and OS, patients were stratified according to whether they had greater or less than 25% residual volume in either the PT or ND at the end of treatment. For this portion of the analysis all patients were evaluated in aggregate and Kaplan–Meier analysis was used to estimate the PFS and OS. Log-rank test was then employed to test the differences in the survival curves.