2017 Section 7 Green Book

North Carolina. 32 This previous cross-sectional study found that at approximately 3 months postdiagnosis, African Americans reported higher physical well-being than non-Hispanic whites. The current study includes the median 3-month survey (baseline) and two addi- tional assessment points at approximately 2 and 3 years postdiagnosis. The strengths of this study relative to other published studies are the long-term HRQOL follow-up assessments, the large population-based study group, and inclusion of a large African American cohort. The overall goal of this study is to identify the sociode- mographic, behavioral, and clinical characteristics asso- ciated with HRQOL over time. A detailed description of the parent study, the Carolina Head and Neck Cancer Study (CHANCE), can be found in Diva- ris et al. 33 Briefly, the CHANCE study—a population-based case-control study of risk factors for HNC—included patients aged 20 to 80 years who were residents of a 46-county region in North Carolina, and who had a newly diagnosed, first primary invasive squamous cell carcinoma between January 1, 2002, and February 28, 2006. Patients were excluded if they had can- cer of the lip, salivary glands, nasopharynx, nasal cavity, and nasal sinuses. In addition, individuals with carcinomas of other histologies, carcinomas at other head and neck sites, or a his- tory of recurrent or second primary tumors were also not eligi- ble. Cases were identified by means of a rapid case ascertainment process in conjunction with the North Carolina Central Cancer Registry in which newly diagnosed cancer cases were identified and reported to the study office every month. This case information was sent to the study office, and initial eligibility was confirmed. With physician’s permission, patients were approached and invited to participate. After consent at a baseline in-person interview, information was obtained on dem- ographics, risk factors, medical history, diet, and other factors. At the baseline interview, permission to obtain medical records was requested. The medical records were collected and abstracted in order to obtain information on each patient’s first course of treatment and pertinent comorbid conditions. All pro- cedures were performed in accordance with the ethical stand- ards of the University of North Carolina at Chapel Hill (Chapel Hill, NC) Institutional Review Board and in accordance with the Helsinki Declaration of 1975, as revised in 1983. A subcohort of CHANCE cases was targeted for a multi- phase prospective assessment of HRQOL. These phases included data collection at up to three time points postdiagno- sis. The first phase of HRQOL data collection (baseline) was conducted in 2005 as part of the overall study baseline in- person interview that collected HRQOL data on cases with a diagnosis of HNC between 2002 and 2005. The median time between diagnosis and the baseline HRQOL data collection was 3 months. Phase 2 (follow-up 1 [FU1]) was conducted using a mailed questionnaire between 2005 and 2008. The median time between diagnosis and the FU1 data collection was 22 months. The third HRQOL data collection phase (follow-up 2 [FU2]) using a mailed questionnaire was conducted between 2007 and 2009. The median time between diagnosis and the FU2 data col- lection was 42 months. A total of 587 patients were included in this longitudinal analysis, of which 133 patients had HRQOL data from all three phases (baseline, FU1, FU2); 37 had base- line and FU1 data; 236 had FU1 and FU2 data; 52 had only baseline data; 127 had only FU1 data; and two had only FU2 MATERIALS AND METHODS Participants and Study Design

data. The response rate was 86% for baseline, 91% for FU1, and 95% for FU2.

Measures The baseline in-person interview collected data on individ- uals’ sociodemographic characteristics, including age, sex, race, education, marital status, and insurance status. Participants reported their health behaviors, including tobacco and alcohol use. Individuals also reported their height and weight (for cal- culating body mass index) and comorbid conditions, including anemia, liver disease, hepatitis, cardiovascular disease, pulmo- nary disease, renal disease, diabetes, and cancers other than head and neck. The comorbid conditions variable was catego- rized as none versus one or more chronic conditions. Patients also self-reported the type of treatment that they received (e.g., surgery, radiation, chemotherapy) and if they were currently using a feeding tube. Health-related quality of life was measured by the Func- tional Assessment of Cancer Therapy–Head and Neck (FACT- H&N) questionnaire. 34,35 The FACT-H&N includes the FACT- General scales of Physical Well-Being (7 questions; score range 0– 28; minimally important difference, [MID] 5 2–3 points), Social Well-Being (7 questions; range 0–28; MID not available), Emo- tional Well-Being (6 questions; range 0–24; MID 5 2 points), and Functional Well-Being (7 questions; range 0–28; MID 5 2–3 points). 36,37 The FACT-H&N includes 12 additional questions on symptoms and issues specific to HNCs, including ability to eat; dry mouth; difficulty breathing; difficulty swallowing; voice qual- ity; ability to communicate; body image; pain in the mouth, throat, or neck; and alcohol and tobacco use. Following the FACT- H&N scoring guidelines, we summed across nine of the 12 ques- tions in the FACT-H&N (excluding alcohol and tobacco use and pain) to create the Head and Neck Cancer Subscale score (score range 0–36; MID 5 3–4 points). All FACT-H&N questions have five response options: “not at all,” “a little bit,” “somewhat,” “quite a bit,” and “very much.” Higher summed scale scores represent better HRQOL. 37 Minimally important differences are defined as the smallest difference in scores between groups that patients perceive as important or personally meaningful, either beneficial or harmful. 37,38 The MIDs were determined in a separate study conducted by the FACT developers using anchor- and distribution-based methods. 37 Statistical Analyses Demographic and clinical characteristics were summarized using proportions. Linear mixed effect models were used on each of the FACT scale scores to account for the longitudinal nature of the data. Time since diagnosis (in months) was used to track time. Each participant may have up to three measure- ments of HRQOL scores, at baseline and follow-up visits 1 and 2. The correlation between any two HRQOL measurements of a participant was specified as exp f 2 d ij = h g , where d ij was the time between the two measurements. This specification reflected stronger correlation between two HRQOL scores when they were measured closer in time and weaker correlation when they were measured farther apart in time. To identify factors associated with the FACT scales, back- ward variable selection procedures were used with stay-in P value set to 0.05. The initial model included time of measurement (lin- ear and quadratic) and all the main effects of the demographic and clinical characteristics listed in Table I, plus their interac- tions with time (linear and quadratic), to allow for the effect of these covariates to vary over time. Among the covariates, feeding tube usage and tobacco/alcohol usage were time-dependent. Once

Reeve et al.: Factors Associated With Quality of Life

130