An Observational Dosimetric Study in Definitively Treated Primary Head and Neck Cancers: To Assess the Effect of Weight Loss and Change in Lateral Neck Dimensions on the Difference Between Dose Planned and Received by the Parotid(s) and Correlation with Adaptive Radiation Therapy

Abstract

Purpose

The head and neck cancer (HNC) literature is rife with reports of differences in planned versus actual radiation doses to the parotid gland (PG) due to changes in anatomy during the course of radiation therapy. We prospectively studied and quantified changes in planned and delivered doses due to weight loss and changes in lateral neck dimensions.

Methods and Materials

Sixty patients were enrolled in this prospective non-randomized observational study. The inclusion criterion was having a newly diagnosed, histologically proven squamous cell carcinoma of HNC. Weight loss (WL) and change in lateral neck dimensions (LND) were assessed weekly, and new hybrid plans were generated using interval replanning CT scans. Dose variations were monitored and extrapolated for replanning CT scans and correlated with WL and change in LND.

Results

The Pearson correlation coefficients for WL and difference in Dmean of ipsilateral and contralateral PG was 0.3292 (P = .0124) and 0.4232 (P = .0010), respectively. There was significantly higher change in the Dmean of bilateral PG (Ipsilateral(I) > contralateral(C)) in patients who experienced WL of >5%. Change in LND correlated with difference in Dmean of ipsilateral PG at 0.4829 (P = .0001) and difference in D50 at 0.4146 (P < .0013). Contralateral PG correlated with difference in Dmean at 0.5952 (P < .0001). The difference in Dmean for ipsilateral PG was 1.8535 Gy for those showing reduction in LND of >1 cm compared with 0.8596 Gy (P = .0091) for those who had ≤1 cm reduction in LND.

Conclusions

Either WL of >5% or reduction in LND of >1 cm can be used as an external parameter to help select patients who might benefit most from replanning and adaptive radiation therapy.

Introduction

Xerostomia is a significant late toxicity of radiation therapy (RT) in head and neck cancer (HNC) patients. Despite the advent of intensity-modulated radiation therapy (IMRT), which has allowed highly conformal therapy and parotid gland (PG) sparing, various studies have shown that the radiation dose planned versus delivered to PG can vary significantly due to tumor response, weight loss, and anatomic changes during RT.¹ This variation in dose to PG occurs because of the decrease in volume and medial displacement of PG during the course of radiation therapy. Although there have been studies correlating variations in planned versus delivered doses to PG with weight loss and changes in neck dimensions, these relationships have not been translated to any meaningful criteria for replanning.²

Therefore, we conducted this prospective study to measure variation in planned versus delivered doses to PG during the course of radiation therapy; we evaluated the correlation of this difference with magnitude of weight loss and change in lateral neck dimensions(LND), both of which can be used as easily measurable external indicators for the need for adaptive radiation therapy.

Methods and Materials

The primary objective of this study was to assess differences in doses planned versus received by PG during radiation therapy (RT) and to correlate the changes in lateral neck dimensions (LND) and weight loss (WL) during RT with dosimetric variations of PG. This was a prospective single-arm observational study. To be included, patients must have been between 18 and 70 years old, had newly diagnosed histologically-proven squamous cell carcinoma of HNC, and had a Karnofsky performance score (KPS) >70. All T stages were allowed. Exclusion criteria were history of any previous irradiation to the head and neck region, any second malignancy/recurrence, primary cancer of PG, collagen disorders, metastatic disease_, nasopharyngeal tumors, lateralized oral cavity tumors, and metastatic neck nodes of unknown origin.

Sample size was calculated at 80% study power and alpha error of 0.05 assuming SD of 1 Gy in cumulative D_mean for dose received by both PG, as found in the study of O'Daniel et al¹

Linear variables are presented as mean and standard deviation (SD) whereas nominal and categorical variables are shown as proportions. Unpaired t tests and Pearson correlation coefficients were used for analysis of linear variables, whereas nominal and categorical variables were analyzed using ANOVA. P < 0.05 was regarded as statistically significant. MedCalc version 16.4 was used for all statistical analysis.

Patients were immobilized in supine position using a customized thermoplastic mask, on an appropriate neck rest. A contrast enhanced CT scan was done with slice spacing of 3 mm from frontal sinus to 2 cm below clavicle head in treatment position on a Philips Brilliance CT Big Bore 16 Slice G-XL-40830 machine. All patients were treated with volumetric arc radiation therapy. Target volumes and organs-at-risk (OAR) definitions were in accordance with International Commission on Radiation Units (ICRU) Report 83 and ICRU Report 50.3, 4 Gross tumor volume (GTV) includes all known gross disease including abnormally enlarged lymph nodes, which can be demonstrated clinicoradiologically. Clinical target volume (CTV) denotes the GTV and subclinical disease (ie, volume of tissue with suspected tumor); planning target volume (PTV) was delineated to provide a margin around each CTV to compensate for the uncertainties of treatment setup and tissue deformation. An isotropic margin of 5 mm was added around the respective high risk and low risk CTV to create the respective high risk (PTV-HR) and low risk (PTV-LR) volumes, which were cropped 5 mm from the surface of the body contour. Dose prescriptions were 60-70 Gy/30-35 fractions at 200-220 cGy/fraction in 6-7 weeks daily, 5 days a week (except on Saturday and Sunday). Dose constraints for OAR were guided by the Quantitative Analysis of Normal Tissue Effects in the Clinic (QUANTEC) guidelines.5, 6, 7 Planning was done on a CLINAC-iX SN4842 linear accelerator and Truebeam STx Varian machine with Eclipse Treatment Planning System Version 15.5. Treatment planning goals are given in Table 1.

Table 1.

Treatment planning goals

PTV	D95% is the minimum absorbed dose that covers 95% of the volume of the PTV.	D95 ≥ 95% of prescription dose
	D2% = minimum absorbed dose received by 2% of volume	D2% ≤ 108% of prescription dose
Parotid	D50% = minimum absorbed dose received by 50% of PG	One or both parotids less than or equal to 30 Gy
	Dmean	Less than 26 Gy
Spinal cord	D2%/Dmax	Less than equal to 45 Gy
Brain stem	D2%/Dmax	Less than equal to 54 Gy
Optic nerve	D2%Dmax	Less than equal to 54 Gy
Lens	D2%/Dmax	Less than 5 Gy

Abbreviations: PG = parotid gland; PTV = Planning target volume.

The initial mean dose (Dmean) and D₅₀ (minimum dose received by 50% of PG) received by PG were calculated and recorded. All patients were monitored weekly for WL and change in LND. A cone beam CT scan was conducted on a weekly basis to evaluate the treatment and any change in PG position over the course of treatment. Also, a re-evaluation planning CT scan was done at 15th fraction (CT1), 20th fraction (CT2), 25th fraction (CT3), and 30th fraction (CT4) (where 33rd or 35th fractions were planned). A hybrid plan was generated by copying the original plan onto the re-evaluation CT scans (CT1,CT2,CT3 & CT4), after proper image registration and matching by aligning the original plan to the treatment isocenter. This allowed us to measure the variation in radiation dose delivered to the PG versus the original plan. Cumulative delivered doses were calculated by averaging the deformed dose distribution for each re-evaluation CT scan.

These changes in original planned doses and the cumulative delivered doses were correlated with WL and changes in LND (calculated at the level of caudal and anterior border of C1 vertebra, which matches the level of PG and is least likely to change with the position of the chin during treatment) occurring over the course of treatment. These dosimetric variables were calculated for each hybrid CT, and their mean values were recorded for each patient (Fig. 1a-b).

Figure 1.

(A) CT0 with original plan and lateral neck dimension (LND) (B) CT1, CT2, CT3, and CT4 super imposed with original CT scan (CT0), with original and LND.

Results

From September 2018 to January 2019, 60 HNC patients were enrolled prospectively, of whom 3 did not complete treatment due to severe toxicities and were thus excluded from the study. As such, 57 patients were considered in the final analysis (Table 2), of whom 24 received 70 Gy in 35 fractions, 13 received 66 Gy in 30 fractions, 2 received 66 Gy in 33 fractions, and 18 received 60 Gy in 30 fractions. A total of 44 (77%) patients had a gap in their radiation therapy schedule ranging from 1 to 13 days with mean gap of 2.23 days.

Table 2.

Patients and tumor characteristics

Median age (years)		55 (Range, 32-69)
Median weight (kg)		70.4 (Range, 43.2-89.8)
T stage	Tx	0
	T1	0
	T2	30
	T3	23
	T4	4
N stage	N0	21
	N1	36
Laterality	Left	30
	Right	27
Male		48
Female		9
KPS	100	0
	90	16
	80	41
Comorbidities
	Diabetes mellitus	15
	Hypertension	10
	Coronary artery disease	1
	None	31

Abbreviation: KPS = Karnofsky performance score.

D_mean of ipsilateral PG was 23.118 Gy and of contralateral PG was 23.445 Gy. Difference in planned and cumulative delivered dose for ipsilateral and contralateral PG were 1.215 Gy and 0.502 Gy, respectively (P < .001). The difference between planned and the cumulative delivered doses was significant for both ipsilateral and the contralateral PG. A total of 55 patients experienced mean WL of 2.38 kg, and 20 patients lost more than 5% of their initial weight. The mean reduction in LND was 8.81 mm, ranging from –1 cm to 4 cm. Seventeen patients showed variation of >1 cm over the course of RT.

Changes in WL and LND were correlated with dosimetric variations in the planned and cumulative delivered doses of PG over the course of radiation therapy. For right PG the Pearson correlation coefficient (r) for difference in Dmean was 0.3798 (P = .0033) and that for D50 was 0.3285 (P = .0118). For left PG the correlation coefficient for difference in Dmean was 0.3827 (P = .003) and that for D50 was 0.3337 (P = .0105). The correlation indicates that WL was directly proportional to the increase in Dmean of PG.

Also, in correlating differences in Dmean and D50 of ipsilateral and contralateral parotids with WL, we found the expected positive relationships. The correlations between WL and difference in Dmean of ipsilateral and contralateral PG were 0.3292 (P = .0124) and 0.4232 (P = .0010) respectively (Fig. 2).

Figure 2.

Pearson correlation coefficient (r) between weight loss and difference in Dmean of ipsilateral parotid gland.

With respect to changes in Dmean for PG for patients who had >5% WL versus those who had ≤5% WL, there was significantly higher change in the Dmean of bilateral PG (ipsilateral(I) > contralateral(C)) in patients who experienced WL of >5%. See Table 3.

Table 3.

Comparisons of differences of parotid Dmeans by weight loss status (unpaired t tests)

Difference	Weight loss	n	Mean	SD	Median	Min.	Max.	P*
D mean	≤5%	36	0.251	0.6771	0.159	−1.205	1.944	.003
C (Gy)	>5%	21	0.932	0.9603	0.98	−0.853	2.879
D mean	≤5%	36	0.921	0.93	0.924	−1.055	2.672	.016
I (Gy)	>5%	21	1.721	1.4932	1.392	−0.922	5.39
D 50 %	≤5%	36	0.195	0.7941	0.0958	−1.323	2.529	.074
C (Gy)	>5%	21	0.647	1.0699	0.534	−0.854	2.894
D 50 %	≤5%	36	0.589	0.9291	0.403	−1.024	2.422	.111
I (Gy)	>5%	21	1.098	1.4486	0.697	−1.476	5.008

Abbreviations: C = contralateral; D_50% = minimum absorbed dose received by 50% of parotid; I = ipsilateral.

^⁎P value of < 0.05 was taken as significant.

Change in LND correlated with difference in Dmean of ipsilateral PG at 0.4829 (P = .0001) and difference in D50 at 0.4146 (P < .0013). Contralateral PG correlated with difference in Dmean at 0.5952 (P = < .0001) and D50 at 0.4978 (P value <0.0001) (Table 4). The data showed a strong correlation between change in LND and change in PG doses with indirect proportionality (ie, as the LND decreased, PG dose increased (Fig. 3a-b)).

Table 4.

Correlation between LND reduction and parotid dose by laterality

	Difference in Dmean ipsilateral	Difference in D50 ipsilateral	Difference in Dmean contralateral	Difference in D50 contralateral
Correlation coefficient (r)	0.4829	0.4146	0.5652	0.4978
Significance level	0.0001	0.0013	<0.0001	0.0001
95% confidence interval for r	0.2544, 0.6604	0.1726, 0.6093	0.3573, 0.7198	0.2726, 0.6713

Figure 3.

(A) Pearson correlation coefficient (r) between difference in lateral neck dimensions (LND_diff) and difference in Dmean of ipsilateral parotid gland (Dmean_diff_ipsilateral) (B) Pearson correlation coefficient (r) between difference in lateral neck dimension (LND_diff) and difference in Dmean of contralateral parotid gland (Dmean_diff_contr).

Regarding differences in Dmean of PG between those who had experienced reduction in LND of >1 cm versus ≤1 cm, significant differences were observed in Dmean for both ipsilateral and contralateral PG. The difference in Dmean for ipsilateral PG was 1.8535 Gy for those showing >1 cm reduction in LND versus 0.8596 Gy (P = .0091) for those who had ≤1 cm reduction in LND. For contralateral PG the mean Dmean increase in patients experiencing >1 cm reduction in LND was 1.3633 Gy versus 0.2203 Gy for those who had ≤1 cm reduction (P = .0011). See Table 5.

Table 5.

Comparison of parotid doses by LND reduction (unpaired t tests)

Difference	LND reduction	n	Mean	SD	Median	Min.	Max.	P*
Dmean	>1 cm	17	1.8535	1.3062	1.507	-0.922	5.39	.0091
(Gy)	≤1 cm	40	0.8596	0.9266	0.8	-1.055	2.85
Dmean	>1 cm	17	1.3633	1.1587	1.043	-0.853	4.473	.0011
C (Gy)	≤1 cm	40	0.2203	0.6943	0.097	-1.205	2.466

Abbreviations: C = contralateral; I = ipsilateral.

* P value of < 0.05 was taken as significant

Discussion

Data analysis revealed that most patients had a significant increase in Dmean of the ipsilateral PG, which was much higher in the patients experiencing >5% WL over the course of radiation therapy. In addition, the patients experiencing >1 cm reduction in LND had a significantly higher cumulative PG Dmean.

We observed that—compared with the contralateral PG—the ipsilateral PG received significantly more dose than planned. The mean of Dmean difference for the ipsilateral PG was 1.215 Gy, which was similar to that reported by O'Daniel et al¹ in their pilot study of 11 HNC patients undergoing radiation therapy by IMRT (a median of 1 Gy increase in Dmean of ipsilateral PG). In another pilot study of 10 HNC patients treated with IMRT, Lee et al⁸ reported >10% increase in Dmean of PG, over that originally planned, in 30% of the patients. In contrast, our study showed >10% increase in Dmean of PG in 11 patients (19%).

Fifty-two (91%) patients experienced WL, of whom 21 (38%) had WL of >5% over the course of radiation therapy. The mean weight loss of the study group was 2.664 kg, ranging from a weight gain of 3.7 kg to a loss of 8.1 kg. The Dmean difference for PG in patients having a weight loss of >5% was significantly higher than for those who had a weight loss of ≤5%, at 1.721 Gy versus 0.921 Gy, respectively, for the ipsilateral PG and 0.932 Gy versus 0.251 Gy, respectively, for the contralateral PG.

There were significant positive correlations (both P's < 0.005) between WL and the Dmean difference for PG planned and delivered with a correlation of 0.3798 for right PG and 0.3827 for left PG. The correlation between WL and difference in D50 difference for right and left PG were also similar, at r = 0.3285 and 0.3337, respectively. Though positively correlated, the correlation was weak at best because increased magnitude of WL did not always align with change in LND due to associated edema. Ahn et al² have reported that WL is positively correlated with increased D50 of PG, observing correlations between 0.30-0.35, which is similar to our findings.

We further analyzed the data for ipsilateral and contralateral PG, expecting to find a stronger correlation between WL and increase in Dmean of ipsilateral PG, but the correlation was similar to what was found in right and left PG individually (r = 0.4829 (P = .0001) for ipsilateral and 0.5652 (P < 0.0001) for contralateral PG).

In our study 50 patients had a reduction in LND by the 25th fraction (ie, after 5 weeks). The mean LND reduction was of 8.81 mm, with a range of –10 mm (ie, an increase in LND due to inflammation/edema) to 40 mm. The mean LND reduction in patients with >5% WL was 13.55 mm, compared with 4.43 mm in those who had ≤5% WL. A total of 50 (87%) patients showed a reduction in LND over the course of radiation therapy. In their study of 23 patients with locally advanced HNC, Ahn et al⁵ reported that mean LND reduction was 14 mm (P < .001), which is comparable to our data in patients who experienced WL >5%. In their study of 15 locally advanced HNC patients undergoing IMRT, Castelli et al⁹ reported a mean LND decrease over the course of radiation therapy to be 7.9 mm, which is comparable to our finding (8.81 mm). They reported an LND decrease in 78% of patients undergoing radiation therapy, compared with 87% found in our study. They also reported an increase of Dmean of PG in 67% of patients, compared with 82% found in our study.

A positive correlation emerged between LND changes and PG dose changes, which was higher than the correlation between WL and difference in PG doses. The correlations between LND reduction and difference in Dmean were 0.4563 for right PG (P = .003) and 0.5391 for left PG (P < 0.0001). Ahn et al² also reported a positive correlation between LND reduction and increase in D50 of PG. However, they reported correlations of 0.17-0.28, which constitute relatively weak associations. In contrast, our study found correlations between LND reduction and increase in D50 of the PG to be 0.5587 (P < .0001) for right PG and 0.5267 (P < .0001) for left PG, which were higher than those observed by Ahn et al.² Lee et al.⁸ also reported a strong correlation between LND reduction and increase in the Dmean of PG. They reported a stronger correlation between increase in Dmean of PG with LND reduction than with WL, which was consistent with what was found in our study. Capelle et al¹⁰ reported change in LND to be the most consistent predictive factor for dosimetric variation in parotid doses. They reported a correlation coefficient of 0.64 (P = .002) between LND reduction and combined mean PG dose. They reported that LND change had a stronger correlation than WL with increase in mean PG dose, which was consistent with our results.

To further quantify the changes seen with LND reduction, we compared Dmean of PG of patients experiencing a reduction of >1 cm versus ≤1 cm and found that Dmean increased significantly for both ipsilateral and contralateral PG at 1.8535 Gy versus 0.8596 Gy, respectively (P = .0091) for ipsilateral and 1.3633 Gy versus 0.2203 Gy, respectively (P = .0011) for contralateral PG. Considering the proximity of the initially planned PG doses to the constraint of 26 Gy, these findings are of clinical importance as even slight variation violates the constraints and can be used appropriately to choose patients for adaptive radiation therapy. To the best of our knowledge this type of quantification for dosimetric variations with change in LND has not been done previously, and this gives us an easily measurable external parameter that can be used to appropriately choose patients for replanning.

Conclusion

Either WL >5% or LND reduction of >1 cm can be used as an external parameter to help select patients who might benefit most from replanning and adaptive radiation therapy.

Limitations

• 1.The current study lacks an appropriate comparison group, which may help quantify the absolute gain in adaptive radiation therapy.
• 2.Instead of continuous variables (daily CT scan and dose calculation), discrete points in RT treatment were used as proxies, which might not represent the true dose deposition.

Disclosures

None.