Treatment Tolerance and Toxicity in Elderly Oropharyngeal Cancer Patients and Implication on Outcomes

PURPOSE/OBJECTIVE(S)

To investigate the tolerance level and toxicity for standard of care treatment for oropharyngeal cancer (OP) in elderly patients and their impact on outcomes.

MATERIALS/METHODS

Using our in-house head and neck cancer database, we looked for non-metastatic OP cases that received definitive treatment between 1/2009-6/2020. All patients received either definitive radiation therapy (RT) +/- concomitant systemic therapy (ST), or surgery followed by adjuvant RT or RT-ST. For the elderly (age at diagnosis ≥65 years) and young (<65 years) patients, we compared treatment package time (TPT) (time from surgery to RT conclusion) for adjuvant RT, total RT duration and unplanned RT interruptions. ST details and dose/protocol modifications were also compared. We evaluated worst grade of pain and mucositis, hospitalization for non-hydration causes and febrile neutropenia (FN) during RT. Feeding tube (FT) use and weight loss were compared. The independent effect of these indicators on locoregional (LRFS), distant (DRFS) recurrence free and overall (OS) survival was assessed using multivariate analyses (MVA).

RESULTS

A cohort of 326 patients was included: 36% elderly (n = 118) and 64% young (n = 208), with no differences in AJCC stage distribution (8th), treatment received and HPV status (HPV+ve: 73% vs 74.6%; p = 0.86). In 23.6 % who received adjuvant RT, median TPT was 86 (range 72-128) and 81 (65-137) days for elderly vs young (p = 0.27); whereas in the definitive RT cases 76.4%, total RT duration was 49 days for both age groups. Overall, prescribed RT course was not completed in 4% and unplanned RT interruptions occurred in 22.8% and both were non-significant between age groups. Among the 261 patients that received ST, elderly utilized more cetuximab (26 vs 12%; p = 0.007). For those who received cisplatin, 20% of elderly had cumulative dose <200 mg/m2 compared to 6% among the younger age group (p = 0.006); and cisplatin was changed to carboplatin or cetuximab in 18% vs 8% (p = 0.019). Delayed/cancelled cycles and dose reductions were similar. There were more hospitalizations (47% vs 27%; p<0.001) and a trend for more FN (9% vs 3%; p = 0.09) with older age, while worst pain and mucositis was similar. FTs were used more in elderly patients (64% vs 50%; p = 0.02), for a median of 97 vs 64 days (p = 0.31); of which 19.5% vs 11% (p = 0.28) were inserted before RT start. However, % weight loss was non-significant. On MVA, longer RT duration, FT use and hospitalizations predicted worse LRFS and DRFS; and they were prognostic for OS in addition to TPT >90 days (p<0.05 for all). Nevertheless, elderly vs young had non-significant 3-year LRFS (91% vs 90% and 67% vs 69%), DRFS (86% vs 90% and 79% vs 81%) & OS (85% vs 81% and 39% vs 52%) for HPV+ve and HPV-ve respectively (p>0.05).

CONCLUSION

Elderly patients with OP need more multi-disciplinary supportive care when receiving RT and concurrent ST. However, survival outcomes are equivalent to younger patients. Ongoing studies should enroll more elderly candidates and stratify endpoints with age.

The Prognostic Impact of Substantial Lymphovascular Space Invasion in Women with FIGO Stage I Uterine Endometrioid Carcinoma with Pathologically Negative Nodal Evaluation

PURPOSE/OBJECTIVE(S)

Substantiallymphovascular space invasion (LVSI) is an important predictor of lymph node involvement in women with endometrial carcinoma. However, its prognostic significance in women with stage I who had pathologic negative nodal evaluation (PNNE) was not fully evaluated. We aimed to evaluate the prognostic significance of substantial LVSI on recurrence-free (RFS), disease-specific (DSS) and overall survival (OS) in women with FIGO stage I uterine endometrioid adenocarcinoma (EC).

MATERIALS/METHODS

Our uterine cancer database was queried for women with stage I EC who had a hysterectomy and PNNE at our institution between 1/1990 and 11/2022. Postoperatively, patients were managed with observation or adjuvant radiation therapy (RT) with pelvic external beam RT or vaginal cuff brachytherapy (VB). Women with synchronous malignancies and those who received adjuvant chemotherapy were excluded. Pathologic specimens were retrieved and LVSI was quantified by Gynecology pathologists (none, focal or substantial). Patients' demographics, surgical and pathologic variables were analyzed. Predictors of RFS, DSS and OS using univariate (UVA) and multivariate analysis (MVA) were studied.

RESULTS

One thousand fifty-two patients were identified with a median age of 63 years and median follow-up of 9.7 years. Median number of examined lymph node (LN) were 9 (range 4-18). 907 patients (86.2%) had no LVSI, 87 (8.3%) had focal and 58 (5.5%) had substantial LVSI. In patients with focal LVSI, 32.2% received pelvic RT and 39.1% received VB. In patients with substantial LVSI, 20.7% received pelvic RT and 58.6% received VB.Recurrence was diagnosed in 86 patients (8.2%). While any LVSI was associated with tumor recurrence, there was no significant difference for the site of initial recurrence between patients with focal vs. substantial LVSI. 5-year RFS was 93.3% (95% CI 91.5-95.1), 76.8% (67.2-87.7) and 79.1% (67.6-95.3) for no, focal and substantial LVSI. The 5-year DSS was 97.6% (96.5-98.7), 83.5% (75-93.1), and 90% (81.8-99.9); and 5-year OS was 90.7% (88.7-92.8), 72.8% (62.9-84.2) and 86% (76.2-97.2), respectively. Independent predictors of worse 5-year RFS and DSS include any LVSI, age > 60 years, higher tumor grade. Independent predictors of worse 5- year OS include any LVSI, age > 60, high comorbidity burden, and higher tumor grade.

CONCLUSION

Our large data suggest similar recurrence-free, disease specific and OS for women with stage I uterine endometrioid carcinoma who had pathologically negative nodal evaluation and substantial or focal LVSI. There was no significant difference for the site of initial recurrence between patients with focal or substantial LVSI. Multi-institutional pooled analyses may be needed to validate our results.

Definitive and Salvage Radiotherapy Compared to Other Modalities for Laryngeal Carcinoma in Situ

PURPOSE/OBJECTIVE(S)

We sought to analyze survival endpoints for laryngeal carcinoma in situ (CIS) undergoing definitive radiotherapy (RT) compared to other modalities.

MATERIALS/METHODS

Usingour prospectively maintained head and neck cancer database, we identified laryngeal CIS patients treated between 6/2001 and 12/2021. We excluded low-grade dysplasia, CIS with any synchronous invasive squamous cell carcinoma (SCC) within 3 months of the initial CIS biopsy and cases with inadequate follow up. Patients were offered either definitive RT, CO2/KTP laser ablation, photodynamic therapy (PDT) or any sort of therapeutic excision. After first line treatment, follow-up includes visits every 3-6 months with laryngoscopy and biopsies as appropriate. For recurrent CIS beyond 6 months of first line treatment, we reported salvage therapies received and long-term outcomes were reported. Using Kaplan-Meier curves and log-rank test we investigated recurrence free (RFS), progression to invasive SCC free (IFS) and overall (OS) survival across treatment groups. Patients managed with salvage RT were compared to first line RT recipients.

RESULTS

Atotal of 85 CIS cases were included: median age 65 years (IQR: 55-74), 73 males (85%) and 70 white (82.4%). 86% had a history of smoking with median pack year of 38 (IQR: 20-55) and 66% had a history of alcohol use. CIS was glottic in most of the cases (90.6%: 66% unilateral, 21% bilateral & 13% involved commissure); with only 9.4% in the supraglottic region. RT was used in 49.4% (n = 42) after biopsy (55%) or surgery (45%) with median dose of 63 Gy/28 fractions, mainly by 3D conformal RT (76%). The remaining 50.6% (n = 43) got therapeutic excision alone (commonly microflap excision) (46.5%), CO2/KTP laser (32.6%) or PDT (20.9%). Demographics and clinicopathological details were non-different between RT and non-RT patients except for Charlson comorbidity index: median 2 (IQR 1-3) in non-RT vs 1 (IQR 0-2) in 1ry RT; p = 0.007. After a median follow-up of 4.8 years (IQR 3.5), 51.8% had recurrent disease, 21.2% progressed to invasive SCC and 9.4% had laryngectomies mainly for invasive SCC after RT. First line RT had improved 2-(83% vs 39%) and 5-(74% vs 22%) year RFS vs non-RT therapies (p<0.001). Nevertheless, 2- and 5-year IFS (89% vs 98% and 80% vs 79%) and OS (92% vs 93% and 81% vs 77%) were non-significant among both (p>0.05 for all). Among non-RT cases with CIS recurrences, 12/35 (34%) had salvage RT. Following RT, salvage RT patients had similar 2- and 5-year RFS (81% vs 83% and 81% vs 74%) and IFS (81% vs 89% and 81% vs 80%) compared to first line RT (p>0.05 for all). All cases with CIS recurrences were salvaged successfully with 100% living with no CIS at latest follow-up.

CONCLUSION

Laryngeal CIS can be treated with a wide range of modalities including 1ry RT which has better recurrence free survival. Nevertheless, non-RT recurrent CIS can be salvaged successfully with many options including RT with equivalent long-term results.

Feasibility of Using Zero-Shot Learning in Transformer-Based Natural Language Processing Algorithm for Key Information Extraction from Head and Neck Tumor Board Notes

PURPOSE/OBJECTIVE(S)

Natural language processing (NLP) technology has the potential to automate information aggregation and summarization in oncology. One example is the automation of patient registry creation. In this work, we aim to show (1) the feasibility of using modern NLP algorithms to extract key information from tumor board notes, and (2) the impact of prompt engineering on the quality of the results.

MATERIALS/METHODS

In this IRB-approved study, we obtained the texts of head and neck tumor board notes for 306 unique patients. Five key pieces of information used to create a patient registry were predefined: age, gender, tumor histology, tumor stage, and primary location. The NLP algorithm used was a modified Text-To-Text Transfer Transformer (T5) model that was initially trained on the Colossal Clean Crawled Corpus (C4) dataset and subsequently fine-tuned on the Stanford Question Answering Dataset (SQuAD) to perform the downstream task of extractive question answering. The NLP model and trained weights were obtained from the Hugging Face platform. During inference, the entire body of the tumor board note and a related question were fed as inputs, and the model predicted a sequence of texts in response to the question. Two sets of questions of similar semantic meanings were used. Questions in prompt set #1 included "What is the gender?", "What is the age?", "What is the type of carcinoma in pathological diagnosis?", "What is the stage?", and "Where is the carcinoma located at?". Questions in prompt set #2 include "Is the patient male or female?", "How old is the patient?", "What kind of cancer?", "What is the cancer stage?", and "What is the tumor location?". Each model-predicted response was compared to the ground truth extracted from the tumor board notes. A response was classified as true if it is consistent with the ground truth, otherwise, it was deemed false. The response accuracy for each question was subsequently calculated.

RESULTS

The median number of words in each tumor board note was 448 (range, 219 - 1505). The accuracy of the NLP algorithm for each question from either set is reported in Table 1. Algorithm performance is higher for extracting objective information such as age, gender, and histology. In addition, it was found that questions of similar semantic meanings but with different wording can lead to significantly different results.

CONCLUSION

We demonstrated that a transformer-based extractive question-answering NLP algorithm can be successfully used for extracting information from head and neck tumor board notes with zero-shot learning. Furthermore, our results highlight the significance of prompt engineering for applying NLP for this task. Future work on finetuning these algorithms to oncology-specific texts can potentially enhance algorithm performance for more difficult tasks.

Disease Characteristics, Treatment and Survival for Oropharyngeal Squamous Cell Carcinoma of Elderly

PURPOSE/OBJECTIVE(S)

Incidence of oropharyngeal cancers (OP) has been increasing over the past few decades, mainly driven by human papilloma virus (HPV) associated cancers in younger men. However, a large number of OP patients in recent years are ≥65 years of age. We wanted to determine if there is a difference in outcomes in elderly patients with OP as compared to younger patients.

MATERIALS/METHODS

We queried our institutional prospectively maintained head and neck cancer database for patients with non-metastatic OP treated between 1/2009-6/2020. We excluded patients who did not receive any definitive treatment. We analyzed clinicopathological and treatment characteristics for elderly (age at diagnosis ≥65 years) compared to young (<65 years) across HPV subtypes. We also studied survival endpoints among age groups using Kaplan-Meier curves and log-rank test. Independent predictors were estimated using multivariate (MVA) Cox regression models for each HPV subtype.

RESULTS

We identified 340 patients who met our inclusion criteria: elderly 123 (36%) and young 213 (64%). The proportion of elderly HPV+ve patients showed an increasing trend over the years studied. Median age was 70 years (range 65-91) in elderly and 56 years (38-64) in young (p<0.001); and HPV+ve/-ve were 73.2/26.8% vs 74.6/25.3% for both age groups respectively (p = 0.86). Elderly patients had higher Charlson Comorbidity Index (CCI) and included more divorcees (p<0.05). There were more elderly current/former smokers (97% vs 82%; p = 0.007) within HPV-ve cases. Definitive radiotherapy (RT) +/- systemic therapy (CRT) was utilized in 73.2% (n = 249), while the remainder had surgery +/- adjuvant RT/CRT. There was no difference with age for OP subsite, 8th edition AJCC stage and treatment received except for more use of cetuximab (22.5% vs 10.2%; p<0.001) and weekly cisplatin (32.4% vs 25.8%; p<0.001) among elderly patients. After a median follow up of 5.24 years (IQR: 3.53), 3-year overall (OS) (HPV+ve: 85 vs 81%; HPV-ve: 39 vs 52%), locoregional free (LRFS) (HPV+ve: 86 vs 90%; HPV-ve: 67 vs 69%) and distant metastasis free (DMFS) survival (HPV+ve: 91 vs 90%; HPV-ve: 79 vs 81%) were all non-significant for elderly vs young respectively. On MVA, CCI and AJCC stage for HPV+ve; and smoking, T-stage and lymphovascular space invasion for HPV- were associated with OS. For HPV+ve, AJCC stage, adjuvant vs definitive RT and treatment in later years were predictive of better LRFS, whereas smoking index and extracapsular space invasion were deterministic for DMFS. Interestingly, outcomes among those who received cetuximab was similar to those who received concurrent cisplatin for all endpoints.

CONCLUSION

We did not note any significant difference in outcomes among elderly patients treated for OP when compared to the younger patients when multi-disciplinary head and neck cancer care is provided. This was noted even though a significantly larger proportion of elderly patients received cetuximab concurrent with RT as opposed to standard of care cisplatin.