Patterns of Symptom Burden During Radiation Therapy or Concurrent Chemoradiation for Head and Neck Cancer: A Prospective Analysis Using the MD Anderson Symptom Inventory–Head and Neck Module

Abstract

Background

We performed a prospective longitudinal study to profile patient-reported symptoms during radiation therapy (RT) or concurrent chemoradiation (CCRT) for head and neck cancer. The goals were to understand the onset and trajectory of specific symptoms and their severity, identify clusters, and to facilitate symptom interventions and clinical trial design.

Methods

Participants in this questionnaire-based study received RT or CCRT. They completed the MD Anderson Symptom Inventory—Head and Neck Module before and weekly during treatment. Symptom scores were compared between treatment groups, and hierarchical cluster analysis was used to depict clustering of symptoms at treatment end. Variables thought to predict symptom severity were assessed using a multivariate mixed model.

Results

Among the 149 patients studied, most (47%) had oropharyngeal tumors, and nearly half received CCRT. Overall symptom severity (P<0.001) and symptom interference (P<0.0001) became progressively more severe and were more severe for those receiving CCRT. On multivariate analysis, baseline performance status (P<0.001) and receipt of CCRT (P<0.04) correlated with higher symptom severity. Fatigue, drowsiness, lack of appetite, problem with mouth/throat mucus, and problem tasting food were more severe for those receiving CCRT. Both local and systemic symptom clusters were identified.

Conclusions

Our findings from this prospective longitudinal study identified a pattern of local and systemic symptoms, symptom clusters, and symptom interference that was temporally distinct and marked by increased magnitude and a shift in individual symptom rank order during the treatment course. These inform us about symptom intervention needs, and are a benchmark for future symptom intervention clinical trials.

INTRODUCTION

Radiation therapy (RT), an essential treatment for head and neck malignancies, is associated with a variety of acute and late toxicities and associated symptoms that affect various normal tissues and organ systems, and interfere with patient activities and functionality. Working groups have emphasized the importance of using patient-reported outcome (PRO) measures rather than relying on physician ratings of toxicity in clinical effectiveness research ^1,2. The use of patient symptom assessment tools has been recommended by the US National Institutes of Health for optimal patient care ³, and are more representative of symptom burden so are now preferred over quality of life (QOL) instruments ¹. Comprehensive routine symptom assessment and effective symptom management are particularly important during the course of RT, with or without concurrent chemotherapy, for head and neck cancer. Unrecognized or uncontrolled symptoms lead to substantial distress and unnecessary suffering. Further, high symptom burden may lead to poorer tolerance for therapy, unplanned treatment interruptions, or modification of the overall treatment schedule, which may negatively affect oncologic outcomes and worsen prognosis ⁴.

Certain acute symptoms resolve after therapy, others may persist and become chronic (“consequential late toxicity”), yet other toxicity-associated symptoms may emerge months or years after the recovery from acute symptoms, and are more prone to increase over time ⁵. It is just over this last decade or so with a shift toward greater primary RT-based treatment and now commonplace CCRT has made us more acutely aware of the potential toxicity costs to patients. However, to our knowledge the pattern of patient-reported local and systemic symptoms during treatment for head and neck cancer has not been fully detailed in a prospective longitudinal study.

The MD Anderson Symptom Inventory is a validated multi-symptom assessment instrument that evaluates patient symptoms commonly associated with cancer and cancer therapy, and the interference with activities of daily living that they cause. Because patients with head and neck cancer have distinct disease site- and treatment-related symptoms, the MD Anderson Symptom Inventory–Head and Neck Module (MDASI-HN) was developed and validated to assess specific head and neck symptoms ⁶, including those mucositis-associated, such as dry mouth, lack of appetite, pain, mouth/throat sores, problems tasting food, problem with mouth/throat mucus, and difficulty chewing/swallowing 7. The MDASI-HN takes about two minutes to complete, and affords clinicians brief, pertinent, and easily attainable patient reported data to help guide patient-specific interventions; to monitor the success of interventions; and to use in comparative outcomes research. We have previously described the head and neck cancer patient symptom profile both before radiation-based treatment has begun ⁸and toward treatment’s end ⁷and modeled patient group-based symptom severity differences and trajectories ⁹, but understanding the timing of onset and trajectory of the severity and resolution of specific symptoms associated with RT or CCRT is necessary for developing programmatic symptom prevention and intervention strategies and for the rational design of symptom-intervention clinical trials.

To address this need, the specific goals of the current study utilizing weekly symptom assessments were to (1) characterize the pattern of acute symptoms, both local and systemic, reported by patients with head and neck cancer during RT or CCRT; (2) evaluate the symptom profiles experienced by selected patient subgroups (RT or CCRT); and (3) identify potential patient-, tumor-, and treatment-related factors associated with symptom severity.

MATERIALS AND METHODS

Patients

This single-institution, prospective questionnaire-based study was conducted after institutional review board approval. Consecutive patients who were able to read and understand English and were scheduled to undergo RT or CCRT for head and neck cancer with curative intent were recruited in our radiation therapy planning and quality assurance clinic ¹⁰. All participants provided study-specific informed consent and were enrolled before any radiation-based treatment was begun. Patient demographic and clinical variables were collected. Patient performance status before radiation-based treatment was rated by the treating radiation oncologist according to the Eastern Cooperative Oncology Group (ECOG) scale ¹¹. Tumor and lymph node status was recorded according to the 6^th (2002) edition of the American Joint Committee on Cancer staging system. Prior surgery and primary tumor site were coded as described elsewhere ¹². Exclusion criteria for this analysis were receipt of treatment with palliative intent, having distant metastatic disease, prior head and neck RT, prior index cancer treatment outside our institution, and failure to complete at least 50% of the cumulative MDASI-HN items across all planned survey time points.

The MDASI–HN assessment tool

The MDASI-HN is a psychometrically validated 28-item questionnaire containing 13 “core” items representing important symptoms common to all cancer types, 9 “head and neck cancer–specific” items, and 6 questions on the extent to which the symptom complex interferes with activities of daily life (“interference” items) ⁶. The MDASI-HN symptom items are rated on numeric 0-to-10 scales from “not present” to “as bad as you can imagine.” The 6 interference items are rated on numeric 0-to-10 scales from “did not interfere” to “interfered completely.” Participating patients self-administered and completed the MDASI-HN questionnaires in hard copy before starting radiation-based therapy (week 0) and then weekly during the 6–7 week course of RT or CCRT.

Statistical analysis

Descriptive statistics were used to summarize patient and clinical characteristics and MDASIHN scores. The overall mean symptom severity and the mean values for each MDASI-HN item over the 6- to 7-week treatment course were plotted for each treatment group (RT vs. CCRT). To assist with clinical interpretation of these data, the proportion of patients with moderate to severe (ratings of ≥5 on the 0-10 scale) and severe (ratings of ≥7) levels of the most common symptoms identified were also calculated, as were symptoms experienced at moderate to severe levels by at least 15% of patients at the end of treatment. Hierarchical cluster analysis of symptoms at the end of therapy was used to provide a pictorial representation of how symptoms clustered at that time. Group mean differences were compared with independent t tests. All reported P values are two-tailed and considered significant if <0.05, unless otherwise noted where Bonferroni corrections were used.

Correlates of higher symptom severity were based on the arithmetic average of the symptom items. Covariates that could influence symptom severity were prespecified and included sex, age, primary tumor site, T category, N category, receipt of previous treatment (surgery or chemotherapy), receipt of CCRT (vs. RT), RT technique (3D conformal or intensity-modulated RT), and baseline (pretreatment) ECOG performance status. We first fitted linear mixed models with only time and a covariate at a time as fixed effects and patients as random effects to determine the independent effect of a covariate on the overall mean symptom severity. Covariates that were significant at P<0.05 were entered as main effects in a multivariate mixed model that allowed us to examine the effect of the predictor variables simultaneously. Analyses were done with IBM SPSS Statistics 21.

RESULTS

Patients, tumors, and previous treatment

Of the 161 patients enrolled in this longitudinal study from 03/2006 to 08/2007, 149 met the current criteria for analysis (77 RT and 72 CRT). The reasons for exclusion were prior radiation to the head and neck (8 patients), treatment with palliative intent (3 patients), and prior outside treatment with limited medical records (1 patient). Patient, tumor, and treatment characteristics are presented in Table 1. The median patient age was 59 years (standard deviation [SD] 11.06). Most tumors (n=116) were of squamous histology. Thirty eight patients received induction chemotherapy; the most common regimens were docetaxel, cisplatin, and fluorouracil (n=15) and a platin and taxane doublet (n=15). Seventy two patients received CCRT; the most common systemic agent was cisplatin (n=38), followed by cetuximab (n=16), multi-agent regimens (n=11), and carboplatin (n=6). Most (92%) of the patients were treated with IMRT; median radiation dose was 66 Gy (range 56-72), delivered in a median of 30 fractions (range 23-42), over a median of 41 days (range 21-52). Most (87%) were treated with standard fractionation radiation treatment schedules.

TABLE 1.

Patient, Tumor, and Treatment Characteristics

	Total (n=149)		Concurrent chemoradiation (n=72)		Radiation therapy (n=77)
	Number	%	Number	%	Number	%
Age
<60 years	79	53	37	51	42	55
≥60 years	70	47	35	49	35	45
Sex
Female	39	26	15	21	24	31
Male	110	74	57	79	53	69
Educational level
12th grade and below	56	38	26	36	30	39
Beyond 12th grade	93	62	46	64	47	61
Ethnicity
White non-Hispanic	122	82	64	89	58	75
Black non-Hispanic	11	7	3	4	8	10
Hispanic	11	7	2	3	9	12
Other	5	3	3	4	2	3
Employment status
Employed outside the home	80	54	38	53	42	54
Homemaker	4	3	3	4	1	1
Retired	48	32	21	29	27	35
Medical leave of absence	6	4	2	3	4	5
Disabled by illness	7	5	4	6	3	4
Unemployed	2	1	2	3	0	0
Other	2	1	2	3	0	0
Disease site
Oropharynx	69	47	42	59	27	36
Oral cavity	18	13	7	9	11	14
Nasopharynx	5	3	5	7	0	0
Larynx	13	9	6	8	7	9
Hypopharynx	6	4	4	6	2	3
Thyroid/Trachea	8	5	0	0	8	11
Major salivary gland	10	7	2	3	8	11
Nasal cavity/paranasal sinus	10	7	3	4	7	9
Skin	4	3	1	1	3	4
Unknown primary site	4	3	1	1	3	4
ECOG performance status
Grade 0	102	69	49	69	53	70
Grade 1	38	26	17	24	21	28
Grade 2	6	4	5	7	1	1
Grade 3	1	1	0	0	1	1
T category
TX/0/1/2	86	59	33	46	53	70
T3/4	49	33	37	52	12	16
Recurrent	12	8	1	1	11	14
N category
NX/0/1	63	43	20	28	43	57
N2	74	51	45	63	29	39
N3	6	4	5	7	1	1
Recurrent	3	2	1	1	2	3
Prior treatment
No prior treatment	71	48	42	58	29	38
Prior chemotherapy	36	24	22	31	14	18
Prior surgery	42	28	8	11	34	44

Pattern of symptom burden during therapy

Regarding assessment compliance, average MDASI-HN completion was 97% in weeks 1-2, 96% weeks 3-4, and 87% weeks 5-6. The mean MDASI-HN item ratings by treatment period and treatment group are depicted in Figure 1 (and Supplementary Table). Overall, the symptom rank order, in terms of symptoms severity, shifted during the treatment course; the 10 symptoms rated as most severe (in order of decreasing mean severity) during weeks 1–2 of therapy were fatigue, dry mouth, drowsiness, problem tasting food, lack of appetite, disturbed sleep, distress, problem with mouth/throat mucus, pain, and problem swallowing/chewing, and during weeks 6–7 they were problem tasting food, problem with mouth/throat mucus, difficulty swallowing/chewing, fatigue, dry mouth, lack of appetite, mouth/throat sores, pain, skin pain/burning/rash, and drowsiness.

Figure 1.

Mean severity of the MD Anderson Symptom Inventory—Head and Neck Module individual symptoms items and overall symptom interference across the treatment period, according to treatment group. Mean individual symptom severity differences for the treatment groups (chemoradiation vs. radiation) for the total sample period were compared by independent t tests; P-values shown (group differences were considered significant if P<0.002 for the 22 symptom items).

Mean weekly severity scores for all 22 symptom items were 0.9 for baseline (pretreatment), 1.2 for week 1, 1.7 for week 2, 2.5 for week 3, 2.8 for week 4, 3.2 for week 5, and 3.4 for week 6 or 7; and the corresponding mean weekly overall symptom interference scores were 1.5 (baseline), 1.5, 1.8, 2.4, 2.8, 2.9, and 3.4. Mixed model results showed that overall symptom severity (P<0.001) and symptom interference (P<0.001) both became progressively worse over the course of treatment. The proportions of patients experiencing moderate-to-severe levels and severe levels of the 13 most prominent symptoms in each treatment period are shown in Figure 2.

Figure 2.

Proportions of patients experiencing moderate to severe levels (ratings of ≥5) (top panel) or severe levels (ratings of ≥7) (bottom panel) of the 13 most commonly experienced symptoms of the MD Anderson Symptom Inventory—Head and Neck Module, according to treatment period.

Results from symptom-cluster analyses at the end of treatment are shown in Figure 3. Most of the top symptoms clustered into one of two distinct symptom groups: the local symptom cluster (e.g., dry mouth, problem with mouth and throat mucus, difficulty chewing/swallowing, mouth/throat sores, and problem tasting food) and systemic symptom cluster (e.g., pain, fatigue, drowsiness, and lack of appetite).

Figure 3.

Dendrogram illustrating the clustering of various symptoms at the end of treatment. Those items that join with others earlier along the relative distance scale of 0–25 (i.e., further to the left in this figure) were rated by patients more similarly. For example, the items pain, fatigue, and drowsiness joined together quickly, indicating that patients perceived and rated these items similarly.

As for differences by treatment group, overall mean symptom severity was worse for those receiving CCRT than for those receiving RT alone (2.9 vs. 1.8, P<0.001) (Fig. 4). The trajectories and comparisons for individual mean symptoms and overall symptom interference over the course of treatment are shown in Figure 1. The five symptoms that were significantly more severe for those receiving CCRT than for those receiving RT were fatigue, drowsiness, lack of appetite, problem with mouth/throat mucus, and problem tasting food (Figure 1). Overall mean symptom interference was higher for the CCRT group than for the RT group (2.5 vs. 2.0, P<0.006).

Figure 4.

Mean overall symptom severity of the MD Anderson Symptom Inventory—Head and Neck Module symptom items across the treatment period, according to treatment group.

Correlates of symptom severity

On univariate analysis, mixed model results showed that receipt of CCRT (P<0.006), T category (P<0.03), and ECOG performance status (P<0.001) were significant correlates of symptom severity over time, whereas sex (P<0.37), age (P<0.47), primary tumor site (P<0.09), previous treatment (P<0.23), N category (P<0.77) and radiation technique (P<0.60) were not. On multivariate analysis, ECOG performance status and CCRT were the only significant correlates of symptom severity: higher mean symptom scores were reported for patients with poorer ECOG performance status (2.4 vs. 2.1, P<0.001) and those receiving CCRT rather than RT (2.5 vs. 2.1, P<0.04). Although symptom severity and group differences by T category were similar in magnitude (2.4 vs. 2.1, P<0.20), the difference was not statistically significant on multivariate analysis.

DISCUSSION

This prospective longitudinal study uses a validated PRO instrument (MDASI-HN) to identify specific symptoms and their trajectory during RT, with or without concurrent chemotherapy, for head and neck cancer. Most patients report relatively low tumor-associated symptom distress before treatment ¹², but this study shows that for 7 of the 22 MDASI-HN individual symptom items, ≥50% of patients report moderate to severe treatment-associated distress at some time during the treatment course (Figure 2 top panel). The specific time of symptom onset and their trajectories inform us about the required optimal timing for specific symptom interventions and patient education and preparedness. Unfortunately, reliable, effective and validated specific interventions for many of these symptoms are not yet available. Nevertheless, our results provide benchmark data for power and sample size estimates for future clinical trials involving symptom interventions.

While we have previously modeled patient group-based symptom severity differences and trajectories ⁹, this study further builds upon those results in that it details the kinetic profile, including the relative emergence and incremental increase in magnitude, for individual local and systemic symptoms and symptom interference, and identifies unique symptoms clusters during RT for head and neck cancer. The most prominent local symptoms clustered: dry mouth, problems with mouth/throat mucus, mouth/throat sores, and difficulty chewing/swallowing. Systemic or global symptoms also formed distinct clusters, and the most prominent of these were fatigue, distress, disturbed sleep, and drowsiness. Relationally, these systemic symptom clusters may be attributed to local cytokine release and the systemic and inflammatory effects of cancer therapy ¹³, as previously demonstrated in other disease sites ^14,15, which may serve as target for future symptom prevention and intervention studies.

Regarding the impact of treatment group on symptom burden, chemotherapy concurrent with RT significantly increased symptom severity. Although this was expected and well known in global terms ¹³, our study provides granular detail. It is notable, for example, that while mouth and throat sores are substantially more severe in the CCRT group, the pain curves for RT and CCRT track relatively more closely. The greatest differences in symptom severity between those receiving CCRT versus RT (figure 3) were seen first early in the treatment course (weeks 1 and 2) and then again, and much more pronounced, toward treatment end (weeks 5 and 6), while severity overlapped between the two groups mid-therapy (weeks 3 and 4). Furthermore, CCRT was associated with significantly increased overall symptom interference and specific interference subscales, and it significantly influenced both the severity and rank order of a subset of both local and systemic individual symptoms, namely fatigue, drowsiness, lack of appetite, problem with mouth/throat mucus, and problem tasting food. We are currently performing studies to evaluate the drivers of interference.

Beyond statistical comparisons, the shapes and trajectories of the overall symptom severity and individual symptom curves help to illustrate symptom differentials between treatment groups. For some symptoms, such as problems with taste, mouth/throat sores, and fatigue, the overall shape of the curves between treatment groups was quite similar, but the curves maintained separation with higher severity of symptoms associated with CCRT. Others, such as dry mouth and mucus, initially track and then diverge mid-treatment; while symptoms such as difficulty with voice/speech, skin pain/burning/rash, and choking track closely without separation. Despite these differing patterns in symptom variation between the RT and CCRT groups, the shape and trajectory of the corresponding interference curves separate at the onset, and remain so throughout.

In terms of the clinical application of the MDASI-HN and the presented results, we believe that weekly sampling, while subject to potential concerns associated with response shift ¹⁶, nonetheless promotes improved treatment compliance and reveals the progressive nature of individual acute symptoms in a manner that allows both prophylactic and reactive interventions to be timed effectively, in a way that maximizes clinic staff allocation (e.g., opportunity for treatment group-specific and/or algorithm-based patient care pathways) and allows patients and caregivers to be given reliable information on the pattern of symptoms to be expected. For example, certain symptoms such as dry mouth, decreased appetite, and problem tasting food appeared early and steadily rose in severity across the treatment course, and as such these symptoms might be emphasized before therapy is begun. Conversely, other common symptoms such as mouth and throat sores and problem swallowing/chewing appeared more abruptly near the midpoint of therapy, with a subsequent rapid rise followed by a relative plateau in severity toward the end of the study period. While patients are also informed of these before treatment, they may benefit from reemphasis mid-treatment. The rise with problems choking in both groups underscores the importance of clinical swallow evaluations, swallowing exercises and maneuvers by speech language pathologists experienced in swallowing therapy, particularly for patients receiving CCRT. Beyond its role as a PRO instrument in multiple ongoing clinical protocols, we are currently using the MDASI-HN as part of our routine patient evaluation and as a management tool before, during and after therapy.

Although these data were prospectively collected, the sampling limitations inherent in a single-institution series (in this case, a tertiary care specialty hospital) may affect the generalizability of our results. However, our study population did reflect recent epidemiologic trends in head and neck cancer, in that most of our patients were younger than 60 years, male, and had primary oropharyngeal cancer. Notably, one would expect many of the tested variables to be correlated with each other. For example, as is true in our recent practice patterns, most patients with advanced nodal disease received induction chemotherapy, most patients with oropharyngeal disease were treated with CCRT (Table 1), and nearly all patients were treated with IMRT rather than 3D conformal RT and standard fractionation radiation schedules were most commonly used. Further, because we excluded patients being treated with palliative intent, most of the patients in this study had good performance status before therapy began. Addressing these limitations will require additional longitudinal studies incorporating additional clinical variables (such as tumor human papillomavirus status, radiation therapy dosimetric details, treatment compliance and objective functional measures) to identify symptom kinetics across a wider array of head and neck cancer patients, treatment modalities, management paradigms, and practice settings.

To our knowledge, this series represents the first report of individual symptom trajectory with symptom interference and cluster analyses based on a validated symptom burden PRO measure, and with high temporal granularity. Many of the studied symptoms do not have current effective, reliable, or validated specific prevention or direct treatment options. Consequently, on the basis of these results, we are conducting multiple prospective, randomized clinical trials evaluating pharmacologic interventions that target purported candidate symptom mechanisms and expression pathways. We are also conducting additional longitudinal analyses to detail the symptom burden after the completion of RT or CCRT so that we can evaluate the influence of these acute symptoms on subsequent late symptom trajectory, recovery, and persistence into long-term survivorship. Additional robust symptom-related data are also necessary for statistical assumptions in future head and neck cancer trials in which symptom distress is an endpoint. This is particularly relevant since recent investigational treatment strategies undertaken by clinical trial cooperative groups have focused on treatment “de-intensification” for selected patients (namely those with human papillomavirus–associated oropharyngeal cancer), with the goals of maintaining or improving established cure rates but reducing treatment-related toxicity and patient symptoms through the added evaluation of trans-oral robotic surgery (Eastern Cooperative Oncology Group Trial 3311 [ClinicalTrials.gov Identifier: NCT01898494]) and comparison of concurrent systemic therapy agents used in definitive chemoradiation strategies (Radiation Therapy Oncology Group Trial 1016 [ClinicalTrials.gov Identifier: NCT01302834]).

In conclusion, most of the patients in our study experienced high symptom severity and symptom burden with a combination of problematic local and systemic symptom constellations. The pattern of symptoms and the degree of symptom interference was temporally distinct, being marked by an increased magnitude and a shift in individual symptom severity rank order over time, and with identifiable clusters of symptom subsets (both local and systemic). These findings will facilitate the rational development of symptom and specifically symptom cluster-based interventions and prevention strategies and help direct the importance of their relative timing. They further provide a reference dataset for study design and statistical assumptions to be used in ongoing prospective clinical trials involving symptom intervention or reduction. Because few reliable interventions have been identified for many of the most problematic local and systemic symptoms, additional research investigating symptom development or expression mechanisms and accompanying targeted symptom reduction strategies are urgently needed.

Condensed abstract (Precis): This prospective, patient-reported symptom assessment study revealed the pattern and severity of both local and systemic symptoms, symptom clusters, and symptom interference during radiation therapy or concurrent chemoradiation for head and neck cancer. These findings will facilitate the rational development of needed clinical studies for symptom and symptom-cluster intervention and prevention.