We utilized iKnowMed™ (iKM) electronic health record (EHR) data maintained by Ontada. iKM is an oncology specific EHR system implemented across The US Oncology Network and select non-network community oncology practices. The US Oncology Network includes 1,400 affiliated physicians operating in over 500 sites of care across the United States and treats approximately 1.2 million cancer patients annually (10).

Study data were primarily sourced from the structured fields of iKM EHR data, with supplemental vital status provided by the Social Security Administration’s Limited Access Death Master File (LADMF). The iKM database captures outpatient encounter histories for patients under community-based care, including (but not limited to) patient demographics (age, race, gender, smoking status), clinical information (disease diagnosis, diagnosis stages, performance status information and laboratory test results), and cancer-directed treatment information (medication, dose, line of therapy, and start and end dates). The LADMF includes records of death reported by family members, funeral homes, hospitals, financial institutions, postal authorities, and federal agencies for persons issued a Social Security card. If there was a conflict between iKM and LADMF vital status data, LADMF data were prioritized because they contain official records of death and should be the most accurate.

The study investigated patients who were diagnosed with locally advanced (stage III – IVB, based on the AJCC 7^th edition) head and neck cancer with histology subtype of squamous cell carcinoma. Study data were extracted from EHRs (i.e., real-world research). This retrospective cohort study included patients aged ≥18 years with de novo newly diagnosed stage III-IVB HNSCC who also initiated CRT; diagnosis and treatment initiation had to be between 01 January 2015 and 31 December 2017 (study period). All eligible patients had a minimum of 2 physician visits during the study period within The US Oncology Network and had data accessible for research purposes. Patients were excluded if enrolled in interventional clinical trials, received treatment indicated for another primary cancer between 01 January 2015 and 31 December 2021, or had salivary gland cancer or nasopharynx cancer (i.e., cancer of nasopharynx, paranasal sinus, or nasal cavity). Given that the study data came from multiple choice/fill-in questions analyzed by a computer, rather than manual analysis of charts (chart review or “unstructured” data), these data are typically referred to as “structured” data in real-world research.

OS, treatment duration, and time to next treatment (TTNT) were assessed. Index date was defined as the first initiation date of CRT received at initial LA HNSCC diagnosis. Surgery was not considered in the TTNT. The iKM EHR structured database contains discrete fields for regimen names, drug names and visit dates. Using the regimen name field, we were able to abstract regimens containing anti-cancer drugs with radiotherapy. Index date was defined as the first initiation date of chemoradiation therapy (CRT) received at initial LA HNSCC diagnosis. TTNT was defined as the interval between index date and initiation of the next treatment or date of death due to any cause. OS was defined as the interval between the index date and the date of death (any cause) as documented in the LADMF or EHR.

Study outcomes were summarized for the overall LA HNSCC population as well as three treatment cohort types: index treatments, primary tumor locations, and HPV status. Index treatment was defined as the first CRT initiated on the index date. Index treatment cohorts included cisplatin + RT, cisplatin + other chemotherapy + RT, and cetuximab + RT. Any systemic anti-cancer drugs other than cisplatin and cetuximab indicated for LA HNSCC were considered as other chemotherapy. Primary tumor location cohorts included cancer of the hypopharynx, larynx, oropharynx, and oral cavity. HPV status included positive, negative, and not documented.

Descriptive methods were used to assess patient demographic, clinical, and treatment characteristics. TTNT and OS were assessed using the Kaplan-Meier method with 95% confidence intervals (CI). In TTNT, patients who did not receive a subsequent treatment and were still alive at the end of the study observation period were censored on the end of the study observation period or the last contact date available, whichever occurred first. In OS, patients who did not have a date of death documented within the study observation period were censored on the end of the study observation period or the last contact date available, whichever came first.

Independent associations of all baseline variables with OS were conducted using a Cox proportional hazard model. All baseline demographic and clinical variables, including age at diagnosis, gender, race, region, stage at diagnosis, primary tumor location, smoking history, body mass index (BMI), Eastern Cooperative Oncology Group performance status (ECOG PS) score, HPV status and index treatments, were included as independent variates in a univariate model. The hazard ratio (HR) and its CI were estimated with ties handled by Breslow (11). For the Cox proportional hazard model building, a type 3 p-value for entry of 0.20 and a type 3 p-value for retention of 0.10 were set a priori. Stepwise selection and clinically relevant covariates were used to build the model.

Cox proportional hazard regression with inverse probability treatment weighting (IPTW) (12), a method to account for differences in cohorts, was used to adjust OS results, given that elderly patients and those with poorer health are more likely to receive cetuximab (13). Multinomial logistic regression was performed to estimate the probability of receiving the treatment regimens of interest, and IPTW was used to reweight and balance cohorts. Extreme values were checked via weight truncation/trimming or weight stabilization if needed. The weights were then used in a Cox regression model to estimate OS. To estimate the variance and CIs of the OS estimates, a robust sandwich variance estimator was used. While no formal cutoff was used for standardized mean difference (SMD) for IPTW analysis (12), the means of values for SMD in IPTW, as well as some general patterns about SMD, were reported.

Figure 1 shows the study population attrition. Initially, 22,782 patients were identified from The US Oncology Network with a documented diagnosis at any time of head and neck cancer. After applying inclusion and exclusion criteria, 1158 patients remained for analysis.

We examined demographics ( Table 1 ) and clinical characteristics ( Table 2 ) at initial diagnosis for the overall study population and stratified by index treatment. The index treatments observed were cisplatin + RT (n=762, 65.8%), cisplatin + other chemotherapy + RT (n=232, 20.0%) and cetuximab + RT (n=164, 14.2%).

Demographic characteristics ( Table 1 ) were generally numerically similar across the treatment cohorts. The cetuximab + RT cohort was numerically older (median 66.6 years), followed by cisplatin + other chemotherapy + RT (62.2 years) and cisplatin + RT (60.8 years). Most patients were male (82.2%) and Caucasian (91.2%). Most patients received treatment at practices in the West (43.7%) and South (31.6%), although the South was the most common for the cisplatin + other chemotherapy + RT cohort (53.9%).

Clinical characteristics were also numerically similar across cohorts, with notable exceptions ( Table 2 ). The cetuximab + RT cohort had the lowest proportion of patients that were overweight or obese (55.8% for cetuximab + RT vs. 64.3% for cisplatin + RT vs. 63.4% for cisplatin + other chemotherapy + RT). The cisplatin + RT cohort had the highest proportion of current tobacco users (45.0%), while former tobacco use accounted for the highest proportion of the cisplatin + other chemotherapy + RT and the cetuximab + RT cohorts (54.1% and 55.1%, respectively). Oropharynx was the most common primary tumor site overall and across index treatment cohorts, found in 775 subjects overall (66.9%). The cisplatin + RT cohort had the numerically highest proportion of patients with ECOG PS score of 0-1 (92.5%), followed by cisplatin + other chemo + RT (84.9%) and cetuximab + RT (79.0%) ( Table 2 ).

Stages at diagnosis, tumor site at diagnosis, and fraction HPV positive were numerically similar among treatment groups. For stages, in the overall group, 25.0% were stage III and 68.7% were stage IVA. TNM classification distribution and tumor location were similar among cohorts. HPV status was positive for the majority of patients overall (69.2%) and for those with oropharynx as the primary tumor site (78.9%). However, HPV status was not documented in the overall study population for 51.8% of the cisplatin + RT cohort, 65.5% of the cisplatin + other chemotherapy + RT cohort, and 54.3% of the cetuximab + RT cohort, and in the oropharynx population for 45.9%, 13.9%, and 45.5% of the same respective treatment cohorts.

Among the treatment cohorts analyzed (cisplatin + RT, cisplatin + other chemotherapy + RT, or cetuximab + RT), cisplatin + RT was the most common concurrent CRT (65.8%) ( Figure 2 ).

For the subsequent treatment ( Figure 2 ) among the overall study population, 838 (72.4%) did not initiate subsequent treatment, 139 (12.0%) died, 96 (8.3%) received non-immune checkpoint inhibitor (immuno-oncology medications, IO)-containing regimens, 46 (4.0%) received nivolumab-containing regimens, and 39 (3.4%) received pembrolizumab-containing regimens.

The median follow-up duration for the overall population was 22.7 months ( Table 3A ). Cisplatin + RT had the longest median follow-up period (25.2 months), followed by cisplatin + other chemotherapy + RT (21.5 months) and cetuximab + RT (12.0 months) ( Table 3A ). The median treatment durations among these three treatment groups were numerically similar (all 1.4 months, data not shown).

Factors associated with OS were analyzed with multivariable (not univariable) Cox analysis. Patients with no documentation were eliminated and a total of 454 patients remained in the final multivariate model. The following covariates were associated with improved survival: never having used tobacco (HR = 0.66; p=0.04), former tobacco user (HR = 0.66, p = 0.01), HPV status of not documented (HR = 0.69 p = 0.05) or positive (HR = 0.32, p<0.0001), overweight BMI (HR = 0.60, P = 0.0039), and obese BMI (HR = 0.54, p=0.002). Covariates associated with poorer survival were age of 61-70 years (HR = 1.67, p=0.001) or 71+ years (HR = 1.8, p=0.004), primary tumor location of hypopharynx (HR = 1.87, p=0.005) or oral cavity (HR = 2.71, p<0.0001) and ECOG PS score of 2+ (HR = 3.27, p<0.0001). Additionally, the results of the univariable Cox model ( Supplementary Table 1 ) are generally in agreement with this multivariable analysis.

TTNT (data not shown) was measured overall and stratified by index treatment. Median TTNT was reached only for the cetuximab + RT cohort (30.4 months). The cisplatin + RT cohort had the numerically highest likelihood of not initiating a subsequent treatment after index at 60 months, followed by the cisplatin + other chemotherapy + RT and the cetuximab + RT cohorts (65.0%, 55.2%, and 39.5%, respectively).

OS ( Figure 3 and Table 3 ) was measured overall and stratified by index treatment ( Figure 3A ), as well as tumor location ( Figure 3B ) and HPV status ( Figure 3C ). As observed with TTNT, median OS was reached only for the cetuximab + RT cohort (54.9 months). The cisplatin + RT cohort had the highest likelihood of survival after index at 60 months, followed by the cisplatin + other chemotherapy + RT and the cetuximab + RT cohorts (74.3%, 60.4%, and 42.8%, respectively).

The mean SMD was 0.141 before IPTW analysis was performed. After applying IPTW, OS was longer in the cisplatin + RT (HR: 0.46; 95% CI: 0.39-0.55) or cisplatin + other chemotherapy + RT (HR: 0.67, 95% CI: 0.58-0.79) compared to the cetuximab + RT cohort. All covariates analyzed via IPTW for these comparisons had a SMD of <0.2 and the majority had a SMD of <0.1 (mean SMD=0.043) (data not shown), indicating the cohorts were balanced (12). The HRs prior to IPTW ( Supplementary Table 1 ) of 0.4 and 0.6 were similar to the results of IPTW of 0.46 and 0.67 for cisplatin + RT and cisplatin + other chemotherapy + RT, respectively, compared to cetuximab + RT.

For primary tumor location ( Figure 3B and Table 3B ), the median follow-up durations/OS in months were 22.7/NR, 15.3/43.2, 14.7/69.8, 13.4/56.8, and 26.8/NR for overall (n=1158), hypopharynx (n=84), larynx (n=164), oral cavity (n=135) and oropharynx (n=775) respectively. The 60-month probability of survival was 67.2%, 35.4%, 59.1%, 47.7%, and 75.1% for overall, hypopharynx, larynx, oral cavity, and oropharynx.

Treatment outcomes ( Figure 3C and Table 3C ) were stratified by HPV status for only the oropharynx subgroup (n=775), in cohorts of HPV positive (n=318), HPV negative (n=71), and HPV undocumented (n=386). Median OS was not reached in any of the HPV cohorts. The 60-month survival rate was greatest for the HPV-positive cohort, followed by the not documented and HPV negative cohorts (87.6%, 67.4%, and 56.4%, respectively).