Intensity modulated proton beam therapy (IMPT) versus Intensity modulated photon therapy (IMRT) for oropharynx cancer patients – a case matched analysis

Abstract

Background

Due to its physical properties, intensity-modulated proton therapy (IMPT) used for oropharyngeal carcinoma patients has the ability to reduce the dose to organs at risk compared to intensity-modulated radiotherapy (IMRT) while maintaining adequate tumor coverage. Our aim was to compare the clinical outcomes of these two treatment modalities.

Methods

We performed a 1:2 matching of IMPT to IMRT patients. Our study cohort consisted of IMPT patients from a prospective quality of life study and consecutive IMRT patients treated at a single institution during the period 2010–2014. Patients were matched on unilateral/bilateral treatment, disease site, HPV status, T and N stages, smoking status and receipt of concomitant chemotherapy. Survival analyzes were performed using a Cox model and binary toxicity endpoints using a logistic regression analysis.

Results

Fifty IMPT and 100 IMRT patients were included. The median follow-up time was 32 months. There were no imbalances in patient/tumor characteristics with the exception of age (mean age of 56.8 years for IMRT patients and 61.1 years for IMPT patients, p-value = 0.010). Statistically significant differences were not observed in overall survival (hazard ratio (HR) = 0.55; 95% confidence interval (CI): 0.12–2.50, p-value = 0.44) or in progression free survival (HR = 1.02; 95% CI: 0.41–2.54; p-value = 0.96). The age-adjusted odds ratio (OR) for the presence of a Gastrostomy (G)-tube during treatment and at 3 months post-treatment are respectively (OR = 0.53; 95%CI: 0.24–1.15; p-value = 0.11) and (OR = 0.43; 95%CI: 0.16–1.17; p-value = 0.10). When considering the pre-planned composite endpoint of grade 3 weight loss or G-tube presence, the odds ratios at 3 months and 1 year were respectively (OR = 0.44; 95%CI: 0.19–1.0; p-value = 0.05) and (OR = 0.23; 95%CI: 0.07–0.73; p-value = 0.01).

Conclusion

Our results suggest that IMPT is associated with reduced rates of feeding tube dependency and severe weight loss without jeopardizing outcome. Prospective multicenter randomized trials are needed to validate such findings.

Introduction

The prognosis of oropharyngeal cancer has improved of the last decades, especially in terms of locoregional control and overall survival, likely due to the increased proportion of Human Papillomavirus (HPV) related tumors. It is now widely accepted that HPV infection is a major causal factor for oropharyngeal carcinoma, especially in non-smoking non-drinking patients [1–3]. It is responsible for the increase in oropharyngeal cancer incidence that is observed worldwide, and notably in North America and Europe. HPV-positive oropharyngeal cancer (OPC) patients are usually younger, have less comorbidities, and more often present with lower T-stages but advanced N-stages [4], and have an improved prognosis compared to HPV-negative patients [5].

Radiotherapy, with or without chemotherapy, is the treatment of choice for most patients with early [6,7] and advanced [8–10] OPC because it allows for organ preservation and avoidance of the morbidities associated with surgical procedures. Avoiding long term sequelae of radiation or chemoradiation is particularly important for patients with OPC as the combination of younger HPV-positive patients with improved disease control outcomes means survivors have the potential to live with the side-effects and complications of treatment for up to 50 years or more. Because it maintains dose levels to the tumor, this strategy could be of interest in all OPC tumors irrespective of HPV status.

Proton therapy, because of its intrinsic physical properties, has the ability to reduce the integral dose delivered to the patients while maintaining a highly conformal target coverage. Dosimetric studies have shown that intensity modulated proton therapy (IMPT) allowed for a dose reduction for various normal tissue structures, including the contralateral submandibular and parotid glands, oral cavity, spinal cord and brainstem, as well as the volume of normal tissue receiving doses of 10, 30, and 50 Gy [11] and in a pediatric population [12,13]. We have previously reported a dosimetric comparison of the first 25 oropharyngeal cancer patients treated with IMPT at our institution and have found that mean doses to the anterior and posterior oral cavity, hard palate, larynx, mandible and esophagus were significantly lower with IMPT compared to generated IMRT comparison plans for the same cohort, as were doses to several central nervous system structures involved in the nausea and vomiting response [14].

Although dosimetric analyses can be hypothesis-generating, analyzing comparative clinical outcomes including safety and efficacy of IMPT relative to photon based IMRT is critical. Therefore, the aim of this study is to report the first case-matched analysis of oropharyngeal cancer patients treated with IMPT or IMRT at a single center during a contemporary period between 2010 and 2014.

Material and Methods

Patient population and matching strategy

From 2011 to 2014, 50 adults OPC patients receiving spot-scanning IMPT with a curative intent were included in an institutional review board–approved observational study where clinical outcomes were prospectively recorded. Participants provided study-specific informed consent. Although tumor outcomes and toxicity data of this population have been reported, a comparative analysis could not be performed [15]. For comparative purposes, IMRT patients were selected from our institutional database which included 512 consecutive patients with oropharyngeal carcinoma treated with IMRT from 2010 to 2012.

IMRT patients were matched with IMPT patients based on factors that impact treatment volumes and expected toxicity during or after radiotherapy. A 2:1 ratio was used to increase statistical power. These factors were, in order: laterality of treatment (unilateral vs bilateral), disease site (tonsil vs base of tongue), p16/HPV status (positive vs negative, missing data being considered as any category), T stage (T1-T2 vs T3-T4), N stage: (N0-N1 vs N2-N3), receipt of concomitant chemotherapy, and smoking status. For smoking status, the cut-off chosen was 5 pack-years (PY, (≤5 vs >5PY) due to difficulty in matching when using the more widely used cut-off of 10 PY. Further matching was attempted on age, but even when using a large age matching range (case age +/− 10 years), the addition of this criterion resulted in the loss of a significant number of patients. It was therefore decided not to match on age but to investigate the age distribution between the two groups and to adjust the toxicity analyses using this factor.

Treatment

The vast majority of OPC cases managed at our institution are treated with a radiation therapy-based approach, and these results have previously been reported [16]. Prior to initiation of therapy, all patients underwent multidisciplinary evaluation within our institution and all cases were presented at our head and neck cancer multidisciplinary tumor board for individualized treatment recommendations regarding sequence and combination of modalities. All patient underwent nutritional counseling and follow-up during and after treatment. Gastrostomy (G-) tube placement was based on a reactive approach. The decision was made after discussion between the patient, treating radiation oncologist and dietician. Reasons for G-tube insertion varied but often included weight loss, inability to maintain oral nutrition and dehydration.

Detailed treatment processes were previously described [15,17,18] and will only be briefly summarized hereafter. All patients underwent non-contrast computed tomography (CT) simulation immobilized in the supine position using full length thermoplastic mask, bite block with or without an oral stent and a posterior customized head, neck and shoulder mold for IMPT patients. During our Head and Neck Radiation Oncology Planning and Development Clinic, all IMRT and IMPT patients were examined by at least two radiation oncologists and target volumes were peer-reviewed for quality assurance purposes [19]. Gross tumor plus margins were prescribed a dose of 66 Gy for small volume disease and 70 Gy for more advanced disease, while the elective regions received 54 to 63 Gy. For IMPT patients, a relative biological effectiveness (RBE) value of 1.1 was used. Carefully selected patients with well-lateralized tonsil cancers underwent ipsilateral neck irradiation [20,21].

IMPT planning was performed with Eclipse proton therapy treatment planning system (version 8.9, Varian Medical Systems, Palo Alto, California). Typically 3 beams were used for whole-field bilateral neck IMPT plans: a left and right anterior oblique and single posterior beam. Multi-field optimization was accomplished for bilateral treatments while single-field optimization was used for unilateral cases. The robustness of each treatment plan was also considered in order to evaluate the sensitivity to uncertainties associated with variations in patient setup and proton beam range in each patient [22,23]. Plan-specific quality assurance measurements were made prior to treatment delivery [24]. Daily kilovoltage image guidance was used for all patients. Verification CT simulation were performed at week 1 and 4 of therapy and adaptive re-planning was considered if inadequate doses were delivered to the targets or the organs at risk.

IMRT planning was performed with Pinnacle planning system (Philips Medical Systems, Andover, MA). Treatment was delivered with a static gantry approach. The template for patients treated to both sides of the neck used 9 beams set equidistant through 360 degrees. Patients treated to only one side of the neck were planned with a template using 7 beams equidistant through a 190 degrees arc. Beam angles and number were modified during the optimization process. In general, IMRT was used to treat the primary tumor and upper neck nodes, whereas the lower neck below the isocenter was treated with an anterior beam, with a larynx and/or full midline block. A “whole-field” IMRT approach was used in situations in which the patient’s anatomy or primary tumor location created concerns that tumor might be under-dosed using the “split-field” approach. IMRT was delivered using Varian (Varian Medical Systems, Palo Alto CA) linear accelerators delivering 6-MV photons with daily image guidance [18]. No systematic re-planning was performed for IMRT patients. The appropriate ICRU recommendations were followed [25,26].

Data collection and endpoint definition

All data were prospectively recorded for the IMPT cohort and retrospectively collected for the IMRT cohort. The data collected consisted of baseline patient and tumor characteristics including: smoking status (pack-years), comorbidities according to the Charlson comorbidity index [27], p16 status, tumor outcomes, acute/late toxicity including emergency room visits, and unplanned hospitalization. Patient weight, placement of a G-tube during or after radiation therapy, patient-rated fatigue or dry mouth (both of which were on a 0–3 scale from none to severe) were recorded prospectively in the radiation oncology medical records at each visit making the coding process more reliable and reduced the rate of missing data. Less than 5% missing data was observed for all endpoints with the exception of weight loss and fatigue at one year post-treatment. Toxicity grades considered are peak grade during radiotherapy and persistent toxicity at three months and one year post-treatment. Toxicity at 2 years post-treatment were collected but not analyzed due to the rate of missing data resulting from the relatively short follow-up of the patients.

Survival times were calculated from the end of radiotherapy to the date of the first event of interest. Events were defined as follows: death from any cause for overall survival (OS), death from any cause or disease recurrence for progression free survival (PFS), and locoregional or distant recurrences respectively for locoregional control (LRC) and distant control (DC). Patients were censored at their last follow-up date.

Statistical analysis

Follow-up was calculated by inverting the censoring indicator and applying the Kaplan-Meier method [28]. The distribution of categorical variables between IMPT and IMRT patients were compared using a chi-square test. Comparisons of the survival distributions between IMPT and IMRT were performed using a log-rank test. Survival curves and estimates of survival at specific time points were computed using the Kaplan Meier method. Multivariable survival analyses were performed using Cox regression which included all variables with p<0.30 in univariable analysis, along with the case (IMPT)/control (IMRT) status. The toxicity endpoints (including the composite endpoint of G-tube and grade 3 weight loss) and the statistical analysis plan were predefined before the statistical analysis. Toxicity rates are reported as the frequency and percentage of patients with toxicity information at a specified time point. IMPT and IMRT patients were further compared using a multivariable logistic regression which included age dichotomized at 60 years as a covariate. Duration of G-tube placement was compared between IMPT and IMRT patients with a logrank test using the removal of feeding tube as an event and censoring patients who died with their feeding tube. All p-values reported are 2-sided and a p<0.05 was considered statistically significant. Statistical analyses were conducted using SAS software (Release 9.3; SAS Institute, Cary, NC, USA).

Results

Patient, tumor and treatment characteristics are presented in Table 1. There were no imbalances between the two groups in any covariate apart from patient age (p-value = 0.01). Median age (range) was 61 years (37–84) for IMPT patients and 55.5 years (34–78) for IMRT patients. Patients had few comorbidities (Charlson comorbidity index of 0–1 in 89.3%). Tumors were in majority locally advanced (N2-N3 in 80%). Most tumors were HPV positive, with only three HPV-negative tumors and 16 patients with an unknown HPV status. Approximately 43% of the patients received induction chemotherapy while two third received concurrent chemotherapy. Unilateral radiotherapy was delivered to 20% of the patients.

Table 1.

Patient, Tumor and Treatment characteristics

Patient characteristics		Entire cohort, n (%)	IMPT, n (%)	IMRT, n (%)	p
Age	≤ 60 years	90 (60)	23 (46)	67 (67)	0.01
	> 60 years	60 (40)	27 (54)	33 (33)
Gender	Female	22 (14.7)	8 (16)	14 (14)	0.74
	Male	128 (85.3)	42 (84)	86 (86)
Tobacco status	0 Pack-Years	70 (46.7)	25 (50)	45 (45)	0.35
	0–10 Pack-Years	21 (14)	4 (8)	17 (17)
	> 10 Pack-Years	59 (39.7)	21 (42)	38 (38)
Charlson Comorbidity Index	0–1	134 (89.3)	45 (90)	89 (89)	0.90
	≥ 2	16 (10.7)	5 (10)	11 (11)
Tumor Site	Tonsil	81 (54)	27 (54)	54 (54)	1.00
	Base of Tongue	69 (46)	23 (46)	46 (46)
P16	Positive	131 (87.3)	44 (88)	87 (87)	0.98
	Negative	3 (2)	1 (2)	2 (2)
	Unknown	16 (10.7)	5 (10)	11 (11)
T Stage	T1-T2	120 (80)	40 (80)	80 (80)	1.00
	T3-T4	30 (20)	10 (20)	20 (20)
N-Stage	N0-N1	30 (20)	10 (20)	20 (20)	1.00
	N2-N3	120 (80)	40 (80)	80 (80)
Induction Chemotherapy	Yes	64 (42.7)	20 (40)	44 (44)	0.64
	No	86 (57.3)	30 (60)	56 (56)
RT Laterality	Bilateral	120 (80)	40 (80)	80 (80)	1.00
	Unilateral	30 (20)	10 (20)	20 (20)
Concurrent Chemotherapy	Yes	96 (64)	32 (64)	64 (64)	1.00
	No	54 (36)	18 (36)	36 (36)
	Cisplatin	39 (41)	13 (41)	26 (41)
	Carboplatin	11 (11)	6 (19)	5 (8)
	Cetuximab	42 (44)	11 (34)	31 (48)
	Taxane	4 (4)	2 (6)	2 (3)
Neck Dissection	Not performed	115 (76.7)	41 (82)	74 (74)	0.50
	Pre Radiotherapy	14 (9.3)	3 (6)	11 (11)
	Post Radiotherapy	21 (14)	6 (12)	15 (15)

Matched variables were unilateral/bilateral treatment, disease site, HPV status, T and N stages, smoking status and receipt of concomitant chemotherapy. The cut-off used for matching on smoking status (≤5 Pack-Years) is different from the one presented in the table due to difficulties in matching using the 10 Pack-Years cut-off. For p16 matching, patients with an unknown p16 status could be considered either p16 positive or negative.

Median follow-up time was 32 months (range 2–55) for the entire cohort, and respectively 29 months (range: 8–49) and 33 months (range 2–55) for IMPT and IMRT patients. Due to differences in the inclusion period, the number of living patients censored before 2 years of follow-up after treatment was 21 (42%) in the IMPT group and 13 (13%) in the IMRT group. Twelve patient deaths were recorded, two in the IMPT group and ten in the IMRT group. OS rates at three years were 94.3% in the IMPT and 89.3% in the IMRT groups. Results from the Cox regression are reported in Table 2. In the univariable analysis (UVA) only advanced T-stage and the insertion of a G-tube at the acute phase were associated with decreased OS, with hazard ratios (HR) of 3.1 (95% confidence interval (CI): 0.98–9.8, p-value = 0.05) and 6.61 (95%CI: 1.8–24.4, p-value = 0.005) respectively. In the multivariable analysis (MVA) the insertion of a G-tube at the acute phase was the only significant factor affecting OS, with a HR of 4.96 (95%CI: 1.1—23.0, p-value = 0.04). The HR between IMPT and IMRT in multivariate analysis was 0.55 (95%CI: 0.12–2.5, p-value = 0.44).

Table 2.

Univariate and multivariate analyses for overall survival

Patient characteristics		Univariate		Multivariate
		HR (95% CI)	p	HR (95% CI)	p
RT type	IMRT	1		1
	IMPT	0.42 (0.09–1.91)	0.26	0.55 (0.12–2.5)	0.44
Age	≤ 60 years	1		–
	> 60 years	1.6 (0.50–4.89)	0.44	–
Gender	Female	1		–
	Male	1.71 (0.22–13.2)	0.61	–
Tobacco status	0 PY	1		–
	0–10 PY	1.63 (0.3–8.9)	0.57	–
	> 10 PY	1.73 (0.49–6.1)	0.39	–
Charlson	0–1	1		1
Comorbidity Index	≥ 2	3.2 (0.87–11.9)	0.08	3.39 (0.81–14.1)	0.09
Tumor Site	Tonsil	1		–
	Base of Tongue	1.17 (0.38–3.6)	0.79	–
P16	Positive	1		–
	Negative	NA		–
	Unknown	1.61 (0.35–7.4)	0.54	–
T Stage	T1-T2	1		1
	T3-T4	3.1 (0.98–9.8)	0.05	1.36 (0.35–5.4)	0.65
N-Stage	N0-N1	1		1
	N2-N3	2.97 (0.38–23.1)	0.30	1.38 (0.14–13.4)	0.78
Induction	No	1		1
Chemotherapy	Yes	2.66 (0.80–8.8)	0.11	1.96 (0.52–7.4)	0.32
RT Laterality	Bilateral	1		–
	Unilateral	NA		–
Concurrent	No	1		1
Chemotherapy	Yes	3.02 (0.66–13.8)	0.15	1.16 (0.20–6.8)	0.91
Neck Dissection	Not performed	1		–
	Pre Radiotherapy	1.03 (0.13–8.3)	0.98	–
	Post Radiotherapy	2.06 (0.55–7.8)	0.28	–
Acute G-tube insertion	No	1		1
	Yes	6.61 (1.8–24.4)	0.005	4.96 (1.1—23.0)	0.04
3-months post RT weight loss	< 20%	1		–
	≥ 20%	0.86 (0.11–6.8)	0.88	–

NA: not assessed. HR is not estimated for HPV-negative or unilateral RT patients due to the very low number of patients/events in these groups.

Twenty-two events (recurrence or death) were observed, seven in the IMPT group and 15 in the IMRT group, leading to a 3-years PFS rate of 86.4% in the IMPT group and 85.8% in the IMRT group, corresponding to a HR of 1.02 (95%CI: 0.41–2.54, p-value = 0.96). The results from the PFS analysis are presented in Table 3. PFS curves according to treatment group are shown in Figure 1. In both the UVA and MVA, advanced age (HR = 2.70; 95%CI: 1.10–6.90; p-value = 0.04) and the insertion of a G-tube at the acute phase (HR = 3.09; 95%CI: 1.19–8.00; p-value = 0.02) were associated with decreased PFS. T-stage trended towards significance in the UVA (p-value = 0.08) but not in the MVA.

Table 3.

Univariate and multivariate analyses for progression free survival

Patient characteristics		Univariate		Multivariate
		HR (95% CI)	p	HR (95% CI)	p
RT type	IMRT	1		1
	IMPT	1.02 (0.41–2.54)	0.96	1.00 (0.39–2.6)	0.99
Age	≤ 60 years	1		1
	> 60 years	3.08 (1.28–7.41)	0.01	2.7 (1.1–6.9)	0.04
Gender	Female	1		–
	Male	1.43 (0.33–6.12)	0.63	–
Tobacco status	0 PY	1		–
	0–10 PY	0.94 (0.19–4.51)	0.93	–
	> 10 PY	2.14 (0.85–5.37)	0.10	–
Charlson	0–1	1		1
Comorbidity Index	≥ 2	2.35 (0.79–7.04)	0.13	1.83 (0.6–5.8)	0.31
Tumor Site	Tonsil	1		–
	Base of Tongue	1.03 (0.44–2.4)	0.94	–
P16	Positive	1		–
	Negative	NA		–
	Unknown	1.25 (0.37–4.3)	0.72	–
T Stage	T1-T2	1		1
	T3-T4	2.27 (0.91–5.64)	0.08	1.15 (0.40–3.31)	0.80
N-Stage	N0-N1	1		–
	N2-N3	1.11 (0.37–3.3)	0.85	–
Induction	No	1		–
Chemotherapy	Yes	1.38 (0.60–3.2)	0.45	–
RT Laterality	Bilateral	1		–
	Unilateral	0.57 (0.16–1.92)	0.36	–
Concurrent	No	1		–
Chemotherapy	Yes	1.37 (0.55–3.37)	0.50	–
Neck Dissection	Not performed	1		–
	Pre Radiotherapy	1.12 (0.25–4.96)	0.88	–
	Post Radiotherapy	2.18 (0.83–5.8)	0.11	–
Acute G-tube insertion	No	1		1
	Yes	3.27 (1.39–7.66)	0.006	3.09 (1.19–8.00)	0.02
3-months post RT weight loss	< 20%	1		–
	≥ 20%	0.91(0.21–3.93)	0.90	–

Figure 1.

Progression free survival according to type of ionizing radiation (IMPT vs IMRT).

Overall, 15 locoregional relapses, 5 in the IMPT and 10 in the IMRT group, were observed. Eight distant relapses were observed, 1in the IMPT and 7 in the IMRT group. Three-year locoregional control rates were 91.0% for IMPT patients and 89.7% for IMRT patients. Three-year distant control rates were 97.8% for IMPT patients and 93.5% for IMRT patients. There were no significant differences between the IMPT and IMRT groups with respect to locoregional control (HR=1.03; 95%CI: 0.35–3.02; p-value = 0.96) and distant control (HR=0.33; 95%CI: 0.04–2.74; p-value = 0.30). In the UVA, LRC was significantly decreased by advanced T-stage (p-value = 0.04) and acute G-tube insertion (p-value = 0.03). Likely due to the low rate of events, there was no significant factor in the MVA for both these endpoints.

The association between acute Gastrostomy (G)-tube placement and survival was studied. The PFS curves according to acute G-tube placement are shown in Supplementary Figure 1. Patients receiving a G-tube placement during radiotherapy had a longer history of tobacco smoking (p-value = 0.03), a higher Charlson comorbidity index (p-value = 0.03), more advanced T and N stages (respectively p-value = 0.01 and 0.07), more frequent bilateral treatment (p-value < 0.001), use of induction (p-value = 0.07) and concurrent chemotherapy (p-value < 0.001) and a treatment duration longer by a mean of 2 days (43 vs 41 days, p=0.0002). In this patient population, sensitivity analyses for PFS have been conducted without using acute G-tube placement as a variable and have yielded comparable results as the initial analyses, and notably the absence of statistically significant difference between IMPT and IMRT groups (HR: 0.82; 95%CI:0.32–2.1; p-value=0.67).

There were no significant differences in acute grade 3 or higher dermatitis or mucositis between the IMPT and IMRT patients (p-value = 0.15 and p-value = 0.90 respectively). Toxicity endpoints between treatment groups are described in Table 4. The median duration of G-tube placement was 2.8 months and 4.8 in the IMPT and IMRT patients respectively (p-value = 0.12). The age-adjusted OR for the use of a G-tube during treatment and at 3 months post-treatment were respectively 0.53 (95%CI: 0.24–1.15; p-value = 0.11) during treatment and 0.43 (95%CI: 0.16–1.17; p-value = 0.10) at 3 months. There was a trend toward lower prevalence of grade 3 weight loss at one year after treatment in the IMPT group, with an OR of 0.28 (95%CI: 0.08–1.05; p-value = 0.06). When considering the predefined composite index of G-tube use or grade 3 weight loss, the odds ratios at 3 months and 1 year were respectively 0.44 (95%CI: 0.19–1.0, p-value = 0.05) and 0.23 (95%CI: 0.07–0.73, p-value = 0.01). Patient-reported grade 2 or higher xerostomia at 3 months was reduced in IMPT patients, with an OR of 0.38 (95%CI: 0.18–0.79, p-value = 0.009). There were no differences between the two groups in grade 2 or higher patient-reported fatigue at any time point in the frequency of emergency room visits, nor in unscheduled hospitalization.

Table 4.

Toxicity Analysis for pre-planned endpoints between IMPT and IMRT at various time points, adjusted for patient age (dichotomized at 60 years)

	During RT				3-Months post RT				1 year post RT
Endpoint	IMPT n (%)	IMRT n (%)	OR (95% CI)	p	IMPT n (%)	IMRT n (%)	OR (95% CI)	p	IMPT n (%)	IMRT n (%)	OR (95% CI)	p
G-tube presence	12 (24)	38 (38)	0.53 (0.24–1.15)	0.11	6 (12)	23 (23)	0.43 (0.16–1.17)	0.10	1 (2)	7 (7.8)	0.16 (0.02–1.37)	0.09
Weight loss >20% compared to baseline	–	–	–	–	4 (8.3)	13 (13.5)	0.64 (0.19–2.11)	0.46	3 (6.7)	17 (19.3)	0.28 (0.08–1.05)	0.06
G-tube OR weight loss > 20%	–	–	–	–	9 (18)	34 (34)	0.44 (0.19–1.0)	0.05	4 (8)	22 (24.7)	0.23 (0.07–0.73)	0.01
Patient rated xerostomia grade 2–3	–	–	–	–	21 (42)	60 (61.2)	0.38 (0.18–0.79)	0.009	21 (42)	42 (47.2)	0.63 (0.30–1.33)	0.23
Patient rated fatigue grade 2–3	39 (78)	84 (86.6)	0.49 (0.20–1.23)	0.13	20 (40.8)	34 (36.2)	1.1 (0.53–2.27)	0.80	7 (14.6)	17 (22.1)	0.5 (0.18–1.36)	0.17
Emergency Room Visit	16 (32)	32 (32)	0.95 (0.45–2.0)	0.89	–	–	–	–	–	–	–	–
Unscheduled Hospitalization	10 (20)	21 (21)	0.92 (0.39–2.2)	0.84	–	–	–	–	–	–	–	–

OR represent the risk of toxicity in IMPT patients compared to IMRT patients. OR was not calculated for G-tube presence at one year post treatment because no patient was present in the IMPT group, but a p-value was calculated using a chi-2 Fisher exact test.

Abbreviations: CI, confidence interval; ND, not determined; OR, odds ratio; RT, radiotherapy

Discussion

This comparative clinical study of oropharyngeal cancer patients treated with IMPT or IMRT suggests that IMPT may achieve toxicity reduction while preserving tumor control. Our study showed that IMPT when compared with IMRT achieves similar cure rates, by both overall survival and disease control measures, after a median follow-up of nearly three years. With regards to toxicity, IMPT is associated with a trend toward a reduction of G-tube rates during the acute phase, at three months, and twelve months. When using a pre-planned composite endpoint of G-tube or grade 3 weight loss, IMPT is associated with significantly reduced rates of toxicity at both 3 months and twelve months.

These clinical results are consistent with a recent case-matched analysis of nasopharyngeal carcinoma patients published by our institution [29] and a recent comparative study in patient receiving ipsilateral cervical radiotherapy [30], and the rates of G-tube are within the range of those reported in randomized trials [31]. One unexpected finding in this patient cohort is that acute G-tube placement was a significant adverse prognostic factor for overall and progression free survival. An analysis of G-tube placement demonstrated strong associations with many adverse features, including greater intensity of treatment as measured by bilateral radiation and use of chemotherapy, more advanced disease, and patients who smoke and have a higher incidence of comorbidities. G-tube placement was also associated with a small (mean two days) but statistically significant longer treatment duration, potentially impacting treatment outcome. Therefore, it is likely that G-tube placement is a proxy for advanced patient and tumor status and more intensive treatment rather than a causal factor for future disease progression. Despite this, reducing the need for feeding tube placement during radiotherapy and at the chronic phase could be an important goal to reduce toxicity and improve patient quality of life (QoL) [32].

The limitations of this study are related to the retrospective coding of IMRT patients, which is known to underestimate toxicity compared to prospective coding. However, the main endpoints, G-tube and weight loss, were limited to the use of objective measures that were captured at the time of occurrence. Further, patient reported levels of fatigue and dry mouth were collected on a regular basis and in a consistent fashion during the entire accrual of the cohort. The prospective coding of IMPT patients, which were all included in quality of life evaluation cohorts, ensures high quality data. The practice regarding feeding tube placement at our institution is reactive and involves the patient along with multiple professionals [33]. It might be argued that the use of feeding tube might have been discouraged in IMPT patients, but this would have likely translated into increased weight loss, which is not observed. Besides, and although the placement of a feeding tube might be physician-biased, feeding tube dependency in the long run, at 3 months and one year, is a strong surrogate for chronic swallowing dysfunction and/or functionally limiting taste and salivary impairment with negative implications on quality of life [32,34]. Reducing these late side effects is an important goal, especially given the high survival rates observed in the HPV positive population.

The second limitation is the absence of matching on age. As increasing age is known to be associated with increasing acute and late toxicity [35,36], and especially feeding tube dependence, toxicity analyses were all age-adjusted. However, the IMPT patients were significantly older than their IMRT counterparts, most likely due to easier IMPT approval for older patient due to Medicare coverage. This age difference should favor IMRT patients in terms of toxicity and notably late feeding tube. In our cohort age above 60 years was found to be only associated with an increased patient reported xerostomia level at three months and one year following treatment. Nevertheless, although patients were matched on multiple factors, the presence of unmeasured confounding factors or selection bias cannot be ruled out. This study remains hypothesis-generating and needs a prospective independent validation.

Several outcomes of interest that might differ between treatment groups, either subjective or objective, such as osteoradionecrosis, late mucosal or neuromuscular toxicity or acute nausea, could not be collected rigorously on the IMRT cohort or due to the relative short follow-up of this study, and have not been analyzed. Other important items, such as full patient reported outcomes or utility metrics would be of interest and some of them have been collected in the IMPT patients but were unfortunately not available in the IMRT cohort.

One major unanswered question relates to the selection of patients for proton therapy. It is anticipated, based on predicted toxicity, that a subgroup of patients would have improved swallowing and salivary function with IMPT compared with IMRT [34,37,38]. The present study does not provide an insight into this issue of patient selection, but rather suggests that differences in toxicity pattern can be detected in a non-selected OPC patient population. The allocation of this scarce and expensive resource based on expected patient benefit is a reasonable option that will however require prospective validation.

Indeed, in the current context of debate about the costs associated with technology improvement in radiation oncology, and especially the cost of proton therapy, this report sheds a new light on the potential value of IMPT for the treatment of oropharyngeal cancers. While the delivery of proton therapy is clearly about two to three times more expensive than the delivery of IMRT [39,40], reducing treatment toxicity in the acute, subacute and chronic phases is a major goal for oropharyngeal cancer patients that could lead to increased quality of life, improved job return and decreased healthcare utilization if costs are considered over the entire disease cycle [39]. Value is defined as the outcomes that matter most to the patients divided by the cost of care [41]. Evaluating value therefore requires an investigation of which outcomes matter to patients. It has been shown that late dysphagia was the major correlate of decreased QoL [42], and before and after treatment patients prioritize swallowing abilities among other functional outcomes [43]. Scientific literature about patient preferences in head and neck oncology is scarce outside of larynx preservation [44] but when asked about their preferences, head and neck cancer patients ultimately prioritize survival over QoL [45]. It might be hypothesized, although this remains to be proven, that the reduction of subacute and chronic side effects observed in IMPT patients could reduce some of the costs associated with the delivery of this advanced technique [39].

Designing a comparative study of IMPT versus IMRT in head and neck cancers is challenging. It is widely accepted that patient-reported outcomes (PRO) should be the measure of interest [46,47], but no PRO set is currently accepted as standard by the medical community for head and neck cancer patients, although multiple PRO tools have been developed in this clinical setting. The clinically meaningful difference of these PROs remains to be evaluated. As no literature is available on PROs in head and neck patients treated with proton therapy, no statistical hypothesis or sample size calculation is feasible. In this context, our study suggests that the chronic presence of a feeding tube or a weight loss of over 20% compared to baseline is a reasonable study endpoint that could be meaningful both for patients and physicians. The systematic collection of PROs and objective radiographic swallowing data (via modified barium swallow studies) at baseline, during treatment and follow-up as secondary endpoints would then provide the basis for the evaluation of patient-reported differences between these two treatments and the basis for evaluating the value of IMPT.

In conclusion, this case-match analysis of oropharyngeal patients treated with either IMPT or IMRT suggests, although additional follow-up and patients are needed, that IMPT provides similar tumor control than IMRT, and is able to reduce subacute and late swallowing-related morbidity as measured by the rate of feeding tube dependency and/or severe weight loss. Given their potential consequences in terms of treatment selection, it is essential that our findings are replicated through prospective multicenter trials, ideally prospective such as the ongoing phase 2–3 randomized trial NCT01893307, or using a model based approach as advocated in Europe and incorporate cost effectiveness analysis as well as patient reported outcomes.