Characterizing Cognitive Outcomes Among Long‐Term Survivors of Locally Advanced Head and Neck Cancer

ABSTRACT

Background

Treatment of locally‐advanced head and neck cancer (HNC) involves multiple therapeutic exposures that may directly or indirectly impact cognition.

Purpose

To describe long‐term cognitive impairment in a cohort of HNC survivors and to identify the cognitive domains most commonly affected.

Methods and Materials

This study enrolled patients with locally advanced or recurrent HNC who received radiation therapy (RT) and survived at least 2 years without recurrence. Participants completed a cognitive battery and quality of life measures. Performance in a cognitive domain was considered impaired if T < 35. Logistic regression was used to assess the relationship between treatment exposures and other variables with impairment in each cognitive domain.

Results

The final cohort consisted of 185 participants. The proportion of patients classified with cognitive impairment by domain was 36.8% for fine motor, 32.7% for verbal memory, 27.0% for executive function, 7.6% for vocabulary, and 7.0% for working memory. In the final model for impaired verbal memory, only radiation dose to the medial temporal lobe (OR = 1.06, 95% CI: 1.00–1.12) was significant. Time from RT (OR = 1.12, 95% CI: 1.01–1.24) and fatigue (OR = 3.45, 95% CI: 1.02–11.59) were associated with impaired working memory. Non‐white race (OR = 8.44, 95% CI: 3.23–22.05) and fatigue (OR = 2.68, 95% CI: 1.24–5.81) were associated with impaired executive function. Impaired emotional function and fatigue, but not cognitive battery performance, were associated with perceived cognitive impairment.

Conclusions

This study supports the hypothesis that many long‐term HNC survivors experience cognitive impairment. Higher radiation dose to the medial temporal lobe was associated with lower verbal memory performance, and fatigue was associated with impaired executive function, perceived cognitive abilities, and overall role functioning.

Summary

This cross‐sectional study of long‐term head and neck cancer survivors found that more than half of survivors met criteria for impairment in at least 1 cognitive domain. Higher radiation dose to the medial temporal lobe was associated with lower verbal memory performance, and fatigue was associated with impaired executive function, perceived cognitive abilities, and overall role functioning.

1. Introduction

The number of head and neck cancer (HNC) survivors is rising rapidly due to effective treatment strategies and increasing prevalence of human papillomavirus‐related cancers that are associated with better outcomes [1]. Nearly 500,000 HNC survivors in the United States have specialized healthcare needs that present a growing public health challenge [2]. Current recommendations for HNC survivorship care are focused primarily on physical impairments such as oral and dental problems, swallowing, lymphedema, and fibrosis [3]. Emerging studies suggest that HNC survivors also experience problems in other health domains including social, emotional, and cognitive [4, 5]. Improving our understanding of additional long‐term impairments of HNC survivors is important in order to develop an improved holistic survivorship care model.

Treatment of locally‐advanced HNC involves multiple therapeutic exposures that may directly or indirectly impact cognition. Radiation therapy (RT) may result in portions of the brain exposed to doses which may cause damage [6]. Radiation may also accelerate carotid artery stenosis and other vascular changes and increase the risk of cerebrovascular events or alter cerebral perfusion [7, 8]. Cisplatin, the most common chemotherapeutic agent utilized for HNC treatment, has previously been linked to cognitive changes in survivors of other cancers [9]. Prior studies suggest that cognitive problems are common in HNC survivors, but have had limited sample size and follow‐up [10, 11]. We performed this study to describe the prevalence of long‐term cognitive impairment in a large cohort of HNC survivors and to identify the cognitive domains most commonly affected. We aimed to identify the association between cognitive impairment and demographic characteristics, clinical factors and therapeutic exposures. We also aimed to characterize the association between objective cognitive performance and self‐endorsed cognitive problems, and to explore how cognitive impairment influences functional status.

2. Methods and Materials

2.1. Study Population and Ethical Review

This cross‐sectional study enrolled patients with locally‐advanced or recurrent HNC who had survived at least 2 years from completion of therapy at the University of Alabama at Birmingham and had no persistent or recurrent disease. Study participants were required to have received at least 50 Gy of radiation and to be > 18 years at the time of diagnosis. HNC was defined as any cancer of the upper aerodigestive tract, salivary glands, base of skull, or cervical lymph nodes with unknown primary site. Patients with cutaneous malignancies were eligible if they received RT that included the cervical lymphatics. Patients with a prior history of cognitive impairment (e.g., diagnosis of dementia) or significant neurologic comorbidity (e.g., stroke with cognitive sequelae) were ineligible. Participants with full‐scale intelligence quotient (FSIQ) T < 30, indicating pre‐existing intellectual impairment, were excluded from further analysis. This study was approved by the Institutional Review Board at the University of Alabama at Birmingham and all participants provided written informed consent.

2.2. Neurocognitive Assessment

Participants completed a battery of standardized neurocognitive tests assessing measures of general intelligence (Wechsler Abbreviated Scale of Intelligence), verbal memory (Wechsler Memory Scale Logical Memory [WMS‐LM] and Verbal Paired Associates [WMS‐VPA] subscales), working memory (The National Institutes of Health Toolbox Cognitive Battery (NIHTB‐CB) List Sort task), vocabulary (Boston Naming Test [BNT]), executive function (NIHTB‐CB Flanker and Card Sort tasks), and fine motor control (dominant hand grooved pegboard test [GPT]). Age‐adjusted T‐scores (mean of 50, standard deviation of 10) were calculated for each measure. Patients with T < 35 on a measure were considered to be impaired in the corresponding cognitive domain. For cognitive domains with 2 tasks (verbal memory and executive function), patients were considered impaired if they scored T < 35 on either task.

2.3. Patient‐Reported Outcomes

Perceived cognitive problems were assessed using the Functional Assessment of Cancer Therapy: Cognitive Function (FACT‐COG) form with sub‐scores below 54 classified as subjective cancer‐related cognitive impairment [12]. Functional status and overall quality of life (QoL) were assessed using the European Organization for Research and Treatment of Cancer (EORTC) QLQ‐C30 form. The presence of clinically‐relevant fatigue was defined as a score of ≥ 39 on the EORTC QLQ‐C30 fatigue symptoms scale and the presence of impaired emotional function was defined as a score of ≤ 70 on the emotional function subscale [13]. The role functioning scale from the EORTC QLQ‐C30 was utilized as a measure of functional status using responses to the questions “Were you limited in doing either your work or other daily activities?” and “Were you limited in pursuing your hobbies or other leisure time activities?” were used to calculate the role functioning scale. Missing items on the EORTC and FACT‐COG forms are accounted for in the resulting summary scores.

2.4. Clinical Data and Therapeutic Exposures

Patients were treated in accordance with institutional care pathways and the National Comprehensive Cancer Network Guidelines for HNC [14]. Clinical information and treatment details, including chemotherapeutic agents and cumulative doses, were abstracted from the electronic medical record. For participants who received a portion of their treatment at another facility we retrieved source documentation when possible. All participants received treatment with photons (predominantly intensity modulated RT) and no patients were treated with proton therapy. Mean radiation exposure to specific brain regions of interest (temporal gyri, medial temporal lobe, anterior and posterior temporal lobes, and vermis) was calculated using Varian Eclipse software (Varian Medical Systems, Palo Alto, CA, USA).

2.5. Analytic Methods

Descriptive statistics were used to summarize the characteristics of the cohort, treatment exposures, and results of the neurocognitive battery. Spearman correlation was used to assess for collinearity among variables. Binary logistic regression controlling for FSIQ was used to assess the relationship between specific variables and impairment (T < 35) in each cognitive domain. Multivariable regression was then performed for each cognitive domain, initially including variables with univariable significance (p < 0.15) and then performing backward selection to retain all variables significant at p < 0.1. For analyses which included radiation dosimetry variables the denominator was adjusted to include only participants for whom these data were available. Univariable and multivariable linear regression was utilized to assess the relationship between domain‐specific cognitive performance and the perceived cognitive impairment (PCI) sub‐score of the FACT‐Cog. Associations between sociodemographic variables, concurrent symptoms, and domain‐specific impairment with role functioning were assessed with univariable and multivariable linear regression. No correction for multiple comparisons was performed given the exploratory nature of the study. Analyses were performed using IBM SPSS version 29.0 (International Business Machines Co., Armonk, NY, USA).

3. Results

Between June 2019 and March 2023, 354 potentially eligible patients were identified, 325 were approached, and 205 participated. Five patients were subsequently found to be ineligible, and 15 were found to have FSIQ T < 30 and were excluded from this analysis, yielding 185 evaluable patients. A comprehensive description of the demographics and treatment details is presented in Table 1. Characteristics of the study participants and non‐participants are presented in Supporting Information S1: Table 1. Most patients were male (71.9%) and white (87%); tumor resection was performed for 57.3% and 61.6% received systemic therapy. Mean age was 64.6 years (range: 25.0–88.0) at the time of study participation and the mean time from RT to study was 5.6 years (2.0–28.9 years). Radiation dosimetry was available for 161 (87.0%) participants. There were no missing neurocognitive data or patient reported outcomes (PRO) form scores.

TABLE 1.

Demographics and treatment details of N = 185 study participants.

Patient characteristics
Age at study participation in years
Mean (± SD)	64.6 (± 10.4)
Time from radiation therapy in years
Mean (± SD)	5.6 (± 4.3)
Sex, n (%)
Male	133 (71.9)
Race and ethnicity, n (%)
White	161 (87.0)
Black	22 (11.9)
Asian	2 (1.1)
Primary site, n (%)
Nasopharynx	13 (7.0)
Oropharynx	96 (51.9)
Oral cavity	22 (11.9)
Larynx	22 (11.9)
Paranasal sinuses	2 (1.1)
Salivary gland	16 (8.6)
Other [1]	14 (7.5)
T‐stage, n (%)
T0	2 (1.1)
T1‐T2	94 (50.8)
T3‐T4	78 (42.2)
N/A2	11 (5.9)
N‐Stage, n (%)
N0	46 (24.9)
N1	44 (23.8)
N2	82 (44.3)
N3	7 (3.8)
N/A3	6 (3.2)
Tumor resection, n (%)
Yes	106 (57.3)
Chemotherapy, n (%)
Concurrent cisplatin	83 (44.7)
Concurrent carboplatin	20 (10.8)
Concurrent cetuximab	13 (7.0)
Concurrent other	13 (7.0)
Neoadjuvant cisplatin	12 (6.5)
Neoadjuvant carboplatin	2 (1.1)
Neoadjuvant other	5 (2.7)
Cumulative cisplatin dose a in mg/m2
Mean (±SD)	221.2 (± 75.2)
Mean (±SD) brain radiation dose b in Gy
Cerebellum dose	13.5 (± 9.2)
Temporal gyri dose	5.6 (± 6.6)
Medial temporal lobe dose	5.2 (± 6.2)

^a Available for N = 77 participants.

^b Available for N = 161 participants.

3.1. Objective Cognitive Outcomes

The distribution of the T‐scores across all neurocognitive measures is presented in Figure 1. Performance was poorest for fine motor control (mean T‐score: 38.6, SD: 13.9), executive functioning (mean T‐score: 41.3, SD: 9.4), and verbal memory (mean T‐score: 45.1, SD: 10.9). The proportion of patients classified with cognitive impairment by domain was 36.8% for fine motor control, 32.7% for verbal memory, 27.0% for executive function, 7.6% for vocabulary, and 7.0% for working memory. No domains were impaired for 40.5% of patients, a single domain was impaired for 32.4% of patients, 2 domains were impaired for 16.2% of patients, and 10.8% were impaired in 3 or more domains.

FIGURE 1.

Box plot of age‐adjusted T‐scores. Dashed red line at T = 35 represents the threshold for impairment.

The correlation matrix of candidate predictor variables is presented in Supporting Information S1: Table 2. As expected, the mean radiation exposure to the cerebellum, temporal gyri, and medial temporal lobes was strongly correlated, with r > 0.7 for each pair but no other correlations with r > 0.3 were observed. The results of univariable and multivariable logistic regression to assess impairment in each neurocognitive domain are shown in Table 2. In the final regression model for impaired verbal memory only mean radiation dose to the medial temporal lobe (OR = 1.06 per Gy, 95% CI: 1.00–1.12) was significant. Time from RT (OR = 1.12, 95% CI: 1.01–1.24) and fatigue (OR = 3.45, 95% CI: 1.02–11.59) were associated with impaired working memory. Non‐white race (OR = 11.04, 95% CI: 2.66–45.87) was associated with impaired vocabulary whereas higher N‐stage (OR = 0.21, 95% CI: 0.05–0.87) was protective. Non‐white race (OR = 8.44, 95% CI: 3.23–22.05) and fatigue (OR = 2.68, 95% CI: 1.24–5.81) were associated with impaired executive function. Tumor location (OR = 2.48, 95% CI: 1.24–4.96 for other versus oropharynx) was associated with impaired fine motor control.

TABLE 2.

Demographic and clinical factors associated with domain‐specific cognitive impairment.

Univariable		Verbal memory				Working memory				Vocabulary				Executive function				Fine motor
		p	OR	Lower 95%	Upper 95%	p	OR	Lower 95%	Upper 95%	p	OR	Lower 95%	Upper 95%	p	OR	Lower 95%	Upper 95%	p	OR	Lower 95%	Upper 95%
Time from RT	Cont.	0.563	1.025	0.942	1.115	0.007	1.215	1.056	1.398	0.008	1.18	1.045	1.332	0.02	1.101	1.019	1.189	0.047	1.075	1.001	1.154
Sex	Male vs. female	0.655	0.83	0.365	1.885	0.029	4.316	1.166	15.975	0.201	2.185	0.66	7.227	0.31	1.483	0.696	3.162	0.402	0.743	0.371	1.489
Race	Non‐white versus white	0.927	0.954	0.345	2.639	0.78	1.234	0.283	5.391	< 0.001	0.069	0.017	0.276	0.01	0.251	0.092	0.681	0.439	0.691	0.271	1.761
Primary site	Oropharynx (ref)	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—
	Nasopharynx or paranasal	0.348	1.796	0.529	6.098	0.767	1.412	0.144	13.835	0.026	11.511	1.346	98.442	0.11	2.639	0.797	8.742	0.183	2.125	0.701	6.442
	Other	0.467	0.727	0.309	1.714	0.909	1.081	0.286	4.092	0.002	12.436	2.456	62.98	0.12	1.864	0.856	4.061	0.01	2.483	1.242	4.964
T‐stage	T3‐T4 versus T0‐T2	0.782	0.901	0.429	1.89	0.12	2.758	0.767	9.913	0.965	1.027	0.318	3.31	0.97	0.985	0.485	2.001	0.908	1.037	0.557	1.933
N‐Stage	N1‐N3 versus N0	0.896	0.948	0.429	2.099	0.531	0.668	0.189	2.358	0.005	0.162	0.045	0.578	0.07	0.498	0.236	1.049	0.629	0.846	0.429	1.669
Tumor resection	Yes versus no	0.702	0.867	0.416	1.803	0.17	2.559	0.669	9.787	0.134	2.667	0.74	9.608	0.83	1.083	0.533	2.199	0.778	1.094	0.586	2.042
Cisplatin	Yes versus no	0.398	1.388	0.648	2.972	0.851	1.147	0.275	4.79	0.096	0.257	0.052	1.27	0.24	0.647	0.314	1.329	0.68	0.877	0.47	1.637
Cerebellum	Cont.	0.181	1.028	0.987	1.071	0.26	1.038	0.973	1.108	0.58	1.019	0.954	1.087	0.5	0.985	0.945	1.028	0.516	1.012	0.976	1.049
Temporal gyri	Cont.	0.056	1.055	0.999	1.114	0.413	1.04	0.947	1.141	0.13	1.07	0.98	1.168	0.61	1.015	0.958	1.076	0.863	1.005	0.955	1.057
Medial temporal lobe	Cont.	0.043	1.061	1.002	1.123	0.506	1.033	0.939	1.135	0.068	1.085	0.994	1.184	0.66	1.014	0.954	1.078	0.794	1.007	0.954	1.063
Fatigue	Yes versus no	0.935	1.034	0.458	2.335	0.050	3.282	1.001	10.756	0.739	0.799	0.213	3.01	0.03	2.215	1.079	4.546	0.117	1.731	0.871	3.439
Multivariate
Time from RT	Cont.					0.038	1.116	1.006	1.239
Sex	Male versus Female
Race	Non‐white versus white									< 0.001	11.035	2.655	45.867	< 0.01	8.438	3.229	22.053
Primary site	Oropharynx (ref)																	—	—	—	—
	Nasopharynx or paranasal																	0.183	2.125	0.701	6.442
	Other																	0.01	2.483	1.242	4.964
N‐Stage	N1‐N3 versus N0									0.031	0.211	0.051	0.869
Tumor resection	Yes versus no
Medial temporal lobe	Cont.	0.043	1.061	1.002	1.123
Fatigue	Yes versus no					0.046	3.445	1.024	11.591					0.013	2.678	1.235	5.806

Note: All models controlled for FSIQ.

3.2. Subjective Cognitive Complaints

Over 40% of patients had perceived cognitive impairment (PCI) by FACT‐Cog. Correlations between the FACT‐Cog PCI sub‐score and each of the neurocognitive tests is shown in Supporting Information S1: Table 3. FACT‐Cog PCI was weakly correlated with verbal memory (WMS‐VPA; r = 0.22, p = 0.004) and executive functioning (NIHTB‐CB List Sort, r = 0.18, p = 0.014; NIHTB‐CB Flanker, r = 0.19, p = 0.013; NIHTB‐CB Card Sort, r = 0.27, p < 0.001). The sensitivity and positive predictive value of FACT‐Cog PCI < 54 were 46.3% and 65.4% for predicting impairment in 1+ domain, 46% and 29.4% for predicting impairment in 2+ domains, and 55% and 14.1% for predicting 3+ domains.

Linear regression models (Table 3) identified impaired emotional function (β = −0.17, p = 0.023), fatigue (β = −0.42, p < 0.001), and FSIQ (β = 0.14, p = 0.032) as variables associated with the FACT PCI sub‐score. No domain‐specific impairment on the objective cognitive assessment was associated with the PCI sub‐score.

TABLE 3.

Linear regression of FACT‐Cog perceived cognitive impairment sub‐score.

	Univariable					Multivariate
	p	Beta	B	Lower 95%	Upper 95%	p	Beta	B	Lower 95%	Upper 95%
FSIQ	0.066	0.130	0.212	−0.014	0.437	0.032	0.136	0.1	0.019	0.415
Verbal memory impairment	0.972	−0.002	−0.087	−4.919	4.745
Working memory impairment	0.587	0.036	2.324	−6.197	10.746
Vocabulary impairment	0.569	0.037	2.281	−5.610	10.173
Executive function impairment	0.256	−0.077	−2.718	−7.398	1.983
Motor coordination impairment	0.156	0.092	2.966	−1.140	7.072
Emotional function impairment	< 0.001	−0.407	−5.95	−20.964	−10.521	0.023	−0.170	−5.664	−19.982	−9.628
Fatigue	< 0.001	−0.524	−18.578	−23.043	−14.113	< 0.001	−0.418	−14.805	−12.191	−0.915

3.3. Functional Outcome

Linear regression models of the EORTC QLQ‐C30 role functioning scale are presented in Table 4. Impaired executive function, impaired fine motor, and fatigue were associated with role functioning by univariable analysis, but fatigue (B = −14.81, 95% CI: −12.19–−0.915), (p < 0.001) was the only significant variable in the final multivariable model.

TABLE 4.

Linear regression of EORTC QLQ‐C30 role functioning scale.

	Univariable					Multivariate
	p	Beta	B	Lower 95%	Upper 95%	p	Beta	B	Lower 95%	Upper 95%
Sociodemographics
Age at participation	0.603	−0.039	−0.084	−0.403	0.235
Sex (male versus Female)	0.056	−0.143	−7.005	−14.204	0.195
Race (non‐white versus White)	0.598	−0.04	−2.623	−12.421	7.174
Years since RT	0.971	−0.003	−0.014	−0.771	0.743
Cognition
Verbal memory impairment	0.605	−0.039	−2.084	−10.029	5.862
Working memory impairment	0.584	−0.041	−3.792	−17.440	9.856
Vocabulary impairment	0.320	−0.075	−6.375	−18.980	6.230
Executive function impairment	0.005	−0.209	−10.423	−17.671	−3.176	0.138	−0.89	−4.46	−10.362	1.443
Fine motor impairment	0.013	−0.185	−8.496	−15.210	−1.783	0.092	−0.1	−4.823	−10.438	0.793
Fatigue	< 0.001	−0.593	−31.797	−37.894	−25.699	< 0.001	−0.575	−30.858	−37.059	−24.656

Note: Bold implies statistical significance.

4. Discussion

Cognition is increasingly recognized as an important component of the overall quality of life, functional outcomes, and vocational outcomes among cancer survivors. HNC survivors have rated cognitive problems as highly important, but there is limited information about the specific cognitive domains that are most commonly impaired among long‐term HNC survivors [15, 16]. In this study we described the prevalence of cognitive impairment in HNC survivors who were at least 2 years from completion of RT and found that over half of patients in our cohort were classified as impaired in at least one cognitive domain using a definition of T < 35. Fine motor control, verbal memory, and executive function were the most frequently impaired domains.

Prior studies of cognition among HNC survivors have been heterogenous in terms of methodology, population, and findings. Most prior research has focused specifically on patients with nasopharyngeal carcinoma which may not generalize well to other HNC survivor populations due to differences in etiology, epidemiology, and brain radiation exposure. Welsch and colleagues provided a review of studies published prior to 2014 and found that nearly all had reported at least some degree of cognitive impairment, including some focusing on non‐nasopharyngeal HNC patients [11]. A telephone study of 122 HNC survivors utilized the Telephone Interview for Cognitive Status and found that more than one‐third of patients had evidence of impairment, but a formal objective cognitive battery was not performed to confirm impairment [17].

Some evidence suggests that a component of cognitive impairment may predate treatment. Piai and colleagues reported mild‐to‐moderate impairment in 12.3%–23.2% and moderate‐to‐severe impairment in 4.7%–10.5% of patients in a cohort of recently diagnosed HNC assessed prior to treatment although how the presence of a recently diagnosed, untreated, and likely symptomatic malignancy impacted battery performance is not clear [18]. A study of 80 HNC patients and 40 controls at the Princess Margaret Cancer Center underwent an extensive cognitive evaluation at multiple time points until 2 years after diagnosis [10]. More than one‐third of the patients with HNC had a decline in global cognition, with specific declines noted in the domains of verbal memory (21%), executive function (26%), concentration (12%) and motor dexterity (10%) with cognitive performance generally declining over the study interval. By comparison, our cross‐sectional study defined impairment as age‐corrected T < 35 at a single time point at a mean of 5.6 years from completion of RT and observed impaired executive function in 27% of patients in our cohort and impaired verbal memory in 32.7%. Despite the difference in methodology, both studies find that impairments of executive function and verbal memory are prevalent in HNC survivors, and the current study indicates that they persist for at least 5 years after completion of therapy.

The larger sample size of our study allowed for analysis of factors that were potentially associated with domain‐specific impairment with special attention to the effects of radiation exposure despite the long interval over which participants received treatment. We explored the effect of RT on domain‐specific impairment by calculating the mean radiation dose to prespecified brain regions. The mean radiation dose to the medial temporal lobe was associated with impaired verbal memory, which is consistent with the large body of literature investigating hippocampal injury in patients with brain tumors [19, 20]. A key difference between our study and most other studies investigating brain radiation exposure and prior studies is that our population consisted primarily of non‐nasopharyngeal HNC patients where exposure was primarily limited to the posterior fossa brain structures. Cerebellum dose was a key variable of interest given, but we did not observe an association between radiation dose to the cerebellum and impairment. More investigation in the setting of a larger longitudinal study is needed to better determine how much cerebellum dose contributes to cognitive outcomes in HNC survivors.

The length of time between RT and cognitive testing was associated with impaired working memory in the final regression model. Very few studies of cognitive outcomes in adult cancer survivors have assessed outcomes more than 3 years from diagnosis. There is emerging evidence that other treatment‐related effects such as cardiopulmonary dysfunction that typically develop after longer latencies may negatively impact cognitive function. More research is needed to determine both the trajectory of cognition after HNC treatment and which post‐treatment factors are associated with long‐term impairment. Other factors associated with domain‐specific impairment included race and tumor location. Race was associated with vocabulary, which may be partly due to intrinsic biases within the cognitive tests themselves, due to normative sampling bias [21, 22]. Specifically, vocabulary knowledge tests is well known to have intrinsic racial biases and subsequent results should be interpreted cautiously. Our study was not designed to formally assess for social or racial disparities, but we recognize that disparities in care patterns and survival have been consistently observed among patients with HNC [23]. The association between fine motor impairment and non‐oropharyngeal tumor location is consistent with the many other outcomes that are more favorable for oropharyngeal tumors, but needs further research.

Fatigue is critical to consider when characterizing cognition in cancer survivors since it impacts both neurocognitive test performance and perceived cognitive abilities [24, 25]. In our study, clinically relevant fatigue (score of ≥ 39 on the EORTC QLQ‐C30 fatigue symptoms scale) was reported by 24.3% of the cohort and was associated with working memory and executive function scores that fell into the impaired range. Fatigue was the only independent predictor of the patients' perception of their cognitive abilities. When we examined role functioning, fatigue was the only variable that remained significant in multivariable analysis despite executive function and fine motor coordination showing univariable significance. This observed link between fatigue, cognition, and role functioning suggests that fatigue will also be critical to address in the context of any strategy to improve cognition in HNC survivors.

Another aim of this study was to characterize the concordance between patient‐reported perceived cognitive abilities and objective cognitive evaluation of impairment for HNC survivors. One previous study evaluated the FACT‐Cog among women with breast cancer where a PCI score less than 54 had good discriminant value for predicting cancer‐related cognitive impairment [12]. On the other hand, multiple studies have found relatively poor correlation between objective and self‐reported cognitive functioning [24, 26]. In our study, emotional function and fatigue were the only variables associated with the FACT‐COG PCI score, which is consistent with others' suggestions that perceived cognitive impairment may be more indicative of emotional distress than objective cognitive impairment [24, 26]. Nevertheless, it is important to consider the perceived experiences of the patient when identifying and developing models of care intervention with regard to cognitive outcomes.

5. Clinical and Research Implications

This study suggests that many HNC survivors experience cognitive impairment, with verbal memory and executive function being the most commonly impaired domains. We observed a relationship between temporal lobe radiation dose and verbal memory impairment. This preliminary finding suggests that clinicians may consider patients with higher temporal lobe exposure at increased risk although confirmatory research is needed. Clinicians should also be aware of a possible discordance between perceived impairment and neurocognitive battery performance. If validated, the finding that perceived cognitive problems were more related to fatigue than objective impairment may particularly impact how clinicians address patient reported problems. From a research perspective this study provides an important benchmark for both objective cognitive impairment and perceived cognitive problems which is critical for design of future interventional trials. This study also highlights the ongoing importance of understanding the interrelatedness of fatigue and cognition in cancer survivors.

6. Limitations

The limitations of this study are important to include. While we were able to assess a larger number of patients with longer follow‐up than prior studies, we utilized a cross‐sectional study design and were unable to assess pre‐treatment cognition. We were also unable to track cognitive impairment over time, thus whether the cognitive problems observed in this study predominantly developed in the early survivorship or if later factors also contributed is unknown. Without baseline information, cognitive impairment cannot be attributed to treatment exposures with certainty. Changes in treatment practices also occurred over the period which are difficult to account for. All statistical models controlled for FSIQ as a measure of cognitive reserve, but we recognize that this does not fully account for inter‐individual differences in baseline cognitive abilities. Cognition may be affected by sociodemographic, lifestyle, and other factors such as smoking and alcohol use that were not reliably available for this study but are prevalent among patients with HNC. Additional potential confounders also include unmeasured comorbidities which affect cognition and baseline vulnerability. Genetics may also play a role in cognition after cancer treatment and a follow‐up study to assess genetic variants in this cohort is underway. Given the study participation rate of less than 60%, the potential for participation bias is an important further limitation. Our assessment of brain radiation dosimetry was somewhat coarse due to the larger field‐of‐view required for HNC simulation CTs and more detailed brain regions could have been defined if MRI scans were available.

7. Conclusion

Cognitive problems were prevalent within this cohort of HNC survivors. Verbal memory and executive function were the 2 most commonly impaired domains in this study, although more extensive future research is needed to better characterize these findings, identify contributing factors, and determine how cognition changes over time. Higher radiation dose to the medial temporal lobe was associated with lower verbal memory performance as well as the total number of impaired domains. Fatigue was associated with impaired executive function, perceived cognitive abilities, and role functioning. These results will be useful to guide the next generation of outcomes studies for HNC survivors as well as strategies to preserve or improve cognition after HNC treatment. Simultaneously addressing fatigue is likely to be integral for any strategy aiming to improve functional outcomes in HNC survivors.

Funding

This study was supported by the NIH grant K08 CA237881.

Ethics Statement

This study was approved by the Institutional Review Board of the University of Alabama at Birmingham.

Consent

All participants provided their written informed consent.

Conflicts of Interest

No authors report any conflicts of interest pertaining to this study. Dr. McDonald, Dr. Cardenas, and Dr. Cardan report research funding from Varian Medical Systems unrelated to this submission.

Permission to Reproduce Materials

The authors have nothing to report.

Supporting information

Supporting Information S1

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.