Assessing the use of anti-PD1 monotherapy as adjuvant therapy and determinants of treatment choice in stage III cutaneous melanoma in the US

This study was a non-interventional study including an online physician survey and a retrospective physician panel-based chart review using electronic medical records.

For physicians to be included in the survey, they had to have been a licensed medical oncologist in the US, have seen at least 5 adult patients with stage III melanoma in the past 5 years, have treated patients with melanoma in the adjuvant setting, be willing to consent, and not currently practice in the State of Vermont. A total of 325 physicians were contacted online via physician panel vendor. No physician identification data was available to the investigators. The physician panel was geographically dispersed in the US, and 152 physicians completed the survey.

For the retrospective chart review, medical oncologists from the nationwide panel identified patients with stage III cutaneous melanoma who had a surgical resection. Patients were included if they were ≥ 18 years old with a diagnosis of stage III cutaneous melanoma who had a surgical resection between January 1, 2018, and December 31, 2021, and whose charts had complete substage information and treatment data for at least 6 months before and 6 months after the index date were included for chart review. To avoid selection bias, physicians randomly selected the patients to be included using a predefined selection algorithm (Additional Methods, Additional File 1) in a 4:1 (AT / no AT proportion), with approximately 100 not treated patients. The 4:1 ratio was chosen to ensure that the group of patients who did not receive adjuvant therapy was not underrepresented, and sufficient sample size was available for the analyses. The index date was the date of surgical resection (wide local excision and sentinel LN [SLN] biopsy, or complete LN dissection [CLND]). The pre-index period was the period from initial melanoma diagnosis to the index date. The post-index period was the period from the index date to the end of follow-up or death, whichever occurred first (Supplemental Fig. 1, Additional File 1). Patients were excluded if they participated in a clinical trial before stage III cutaneous melanoma diagnosis, received neoadjuvant and/or adjuvant treatment for stage III melanoma in a clinical trial after diagnosis, or if they were diagnosed with an active second non-in-situ malignancy in the past 3 years prior to index date.

The online physician survey collected data on physician characteristics (i.e., specialty, practice setting, geographic region, years of medical practice, years of treating patients with cutaneous melanoma, number of stage III melanoma patients seen in the past 5 years, proportion of patients with stage III melanoma receiving adjuvant therapy, proportion of patients referred by dermatologists, surgical oncologists, primary care providers, or other specialties), and on physician perceptions about adjuvant treatment of patients with stage III melanoma (i.e., drivers of adjuvant treatment choice, characteristics of the high-risk patient population, treatment guideline referenced, frequency of shared decision making with patients regarding adjuvant therapy use, proportion of patients who refused adjuvant therapy, and reasons why patients refuse adjuvant therapy).

The retrospective chart review study collected information on demographics at resection (i.e., age, sex, race/ethnicity, geographic region, and insurance coverage), clinical characteristics (i.e., comorbidities, disease substage at diagnosis, Eastern Cooperative Oncology Group Performance Status [ECOG PS], pathology data, and laboratory tests, including lactate dehydrogenase [LDH] levels and BRAF mutation status), treatment history before resection, and surgical resection and treatment following resection (i.e., disease substage before resection, initial diagnosis or recurrence before resection, type of lymph node dissection, receipt of radiotherapy, and receipt of steroids after resection).

All data extracted from medical records were entered into the electronic case report form (eCRF). The physician survey was also integrated into the eCRF for data entry by participating physicians. In addition, automated logic and range checks were implemented within the eCRF to help ensure data quality at the point of entry. Before data collection, the eCRF was pretested with two eligible oncologists from the nationwide panel through one-on-one interviews to obtain input on clarity and ease of data entry. The eCRF and survey were reviewed by the physicians and revised based on feedback from these pretests. A “soft launch” of the eCRF and survey was then conducted with 54 patients. The quality of the soft launch data was reviewed manually; this included checking variable ranges and data consistency between related questions. Potential issues (e.g., areas of misinterpretation and data entry errors) were addressed before full launch of the eCRF and survey.

This study did not require patient informed consent, as only deidentified data were collected from the chart review. All study materials were reviewed by the WIRB-Copernicus Group Independent Review Board, which granted an exemption determination on November 29, 2022 per Title 45 of CFR, Part 46.104(d)(2, 4) as no personally identifiable information was collected [7].

Descriptive analyses were used to report the data from the physician survey, patient demographics and clinical characteristics, and treatment utilization. Continuous variables were described using means, standard deviations, and medians; categorical variables were described using counts and percentages. Analyses were stratified by AJCC8 substage (IIIA vs. IIIB, IIIC, or IIID [IIIB/C/D]) and by adjuvant treatment received (i.e., anti-PD1 monotherapy, BRAF/MEK therapy, no treatment) and the subgroup that did not receive adjuvant treatment.

The associations between patient and disease characteristics and AT use were assessed using logistic and multinomial regression models, overall and stratified by AJCC8 substage (IIIA vs. IIIB/C/D). In the logistic model, the dependent variable was whether a patient received AT; in the multinomial model, it was the type of AT received (anti-PD1 monotherapy, BRAF/MEK, or no AT). For both models, the independent variables were patient demographic and clinical characteristics including age, sex, insurance coverage, comorbidities on the index date, ECOG PS after the index date, baseline tumor ulceration, baseline LDH (above upper limit of normal, normal range, etc.), presence of lymphovascular invasion, presence of regression in primary specimen, BRAF mutation status after the index date, disease substage before resection on the index date, type of resection on the index date, receipt of radiotherapy after the index date, receipt of steroids after the index date, deep margin status, and year of surgical resection. Backward selection was used for both logistic and multinomial regressions to trim redundant predictors from the model; the final model chosen was the one with the lowest Akaike information criterion.

All analyses were conducted using RStudio Version 1.2.1335.

A total of 152 medical oncologists who had treated patients with stage III melanoma completed a survey based on their clinical experience and reviewed the charts of 507 patients diagnosed with stage III melanoma whom they had treated. Of these 507 patients, 405 (79.9%) received AT (360 received anti-PD1 therapy [222 pembrolizumab and 138 nivolumab] and 45 received BRAF/MEK therapy).

Physicians most frequently practiced medicine in private practice (40.1%) or in an academic institution (38.2%) and lived in the South (37.5%) or Northeast (25.0%) region. They reported having treated patients with cutaneous melanoma for a mean of 15.3 years, half of them followed > 50 stage III melanoma patients in the past 5 years, and 75.8% of their stage III melanoma patients received AT after resection.

Physicians most frequently reported clinical guidelines (61.2%), treatment efficacy (37.5%), ECOG PS (31.6%), disease stage (21.7%), and patient comorbidities (18.4%) as the key drivers of AT prescription. Most physicians (76.7%) reported that over half of their patients had available BRAF mutation status information after surgery, and 68.0% usually or always considered BRAF mutation status when offering AT to their patients. In addition, the majority (55.9%) reported that the decision to use AT was usually or always shared with patients over half of of the time. A large proportion of physicians (59.9%) reported having had a patient refuse AT. The most commons reasons for patients refusing AT were concerns about treatment-related toxicity (68.1%), expected impact on daily life and daily activities (44.0%), cost/insurance coverage (37.4%), and comorbidities (e.g., autoimmune disease) (36.3%).

The mean age at the time of resection was 58.1 years, 56.0% of the patients were male, most patients were White (76.1%) and commercially insured (59.0%). At the time of resection, patients had substage IIIA (33.1%), IIIB (37.1%), or IIID (23.3%) disease (Table 1).

A total of 71.6% of patients had at least one comorbidity, and 82.0% had an ECOG PS of 0 or 1. The most common comorbidities were hypertension (38.7%), coronary artery disease/myocardial infarction (11.8%), asthma (9.5%), anxiety disorder (8.1%), and depression (7.5%; Table 1). Baseline LDH levels on the date of resection or in the preceding 6 months were obtained for 370 patients. Of these, 231 (62.4%) had normal LDH levels and 139 (37.6%) had levels that were above the upper limit of normal. In addition, BRAF mutation was found in 31.0% of the patients (Table 1). Pathological findings included ulcerations, observed in 52.9% of patients, lymphovascular invasion in 41.6%, and positive deep margins in 22.3% (Table 1).

A total of 81.6% of patients had the primary lesion resected before the main surgical resection. Most patients (92.5%) had a resection following initial melanoma diagnosis, while 6.9% had a resection following disease recurrence. The resection included a complete lymph node dissection among 74.6% of the patients, while 18.9% had sentinel node biopsy only (Supplemental Table 1, Additional File 1). After resection, 26.0% of patients received radiotherapy, either in the primary site (80.3%) or in the lymph node basin (68.9%). About 10.5% of patients received steroids after surgical resection (Supplemental Table 1, Additional File 1).

Among the 507 patients in the study, 71.0% received AT with anti-PD1, 8.9% had BRAF/MEK therapy, and 20.1% did not receive any AT (Fig. 1). Further, patients resected in recent years were significantly more likely to receive anti-PD1 therapy after resection compared to BRAF/MEK or no treatment (Supplemental Fig. 2, Additional File 1). Among patients treated with adjuvant anti-PD1 therapy, 222 (54.8%) received pembrolizumab and 138 (34.1%) received nivolumab (Table 2). The mean time from resection to treatment initiation was 37 days. Median duration of AT was 357 days, 78.4% of patients discontinued treatment, and 25 (6.2%) had a change in dosage or dose frequency (Table 2).

The primary documented reasons in the patient charts for not prescribing AT included expected limited benefits of treatment (37.3%), patient’s refusal of AT (36.3%), concerns about treatment-related toxicity (30.4%), and cost/inadequate insurance coverage (25%; Fig. 2A). The documented main reasons for physicians selecting anti-PD1 therapy were to improve patient outcomes (79.2%), to limit disease progression (18.6%), patient request (13.3%), favorable cost/insurance coverage (7.5%), and high adherence (5.8%; Fig. 2B).

Among all stage III patients, compared with patients who did not receive AT, having disease substages IIIB/C/D significantly increased the probability of receiving AT (odds ratio [OR] = 1.74; 95% confidence interval [CI]: 1.05, 2.89; P < 0.05). Additionally, Medicaid/no insurance (OR = 0.42, 95% CI: 0.21, 0.81; P < 0.05) and ECOG PS of 2/3 (OR = 0.37, 95% CI: 0.20, 0.70; P < 0.01) were associated with a significantly lower probability of receiving AT (Fig. 3).

The mean age at the time of resection was 58.8 years for patients with stage IIIA disease and 57.8 years for patients with stage IIIB/C/D disease; 50.6% among stage IIIA patients and 58.7% among stage IIIB/C/D patients were male (Table 1).

The presence of comorbidities was similar between stage subgroups (69.6% substage IIIA and 72.6% substages IIIB/C/D). Patients with coronary artery disease/myocardial infraction were more frequent in the IIIB/C/D substage (14.2% vs. 7.1% in the stage IIIA subgroup). The proportion of patients with an ECOG PS of 0 was higher in the IIIA subgroup (39.3% vs. 29.8% in the IIIB/C/D subgroup), whereas the proportion with an ECOG PS of 1 was higher in the IIIB/C/D subgroup (52.8% vs. 41.7%; both P < 0.01; Table 1). A significantly higher proportion of patients in the IIIB/C/D subgroup had elevated levels of LDH compared to patients in the IIIA subgroup (45.6% vs. 21.3%). Patients in the IIIB/C/D substages had significantly higher proportions of ulceration, positive margin status, and lymphovascular invasion compared to patients in the IIIA substage (Table 1). Patients with stage IIIA more frequently had a resection of the primary lesion before the main surgical procedure (91.4% vs. 77.3%). No other significant differences were observed in treatment utilization between substages (Supplemental Table 1, Additional File 1).

Improving patient outcomes was a significantly more common reason for selecting anti-PD1 therapy in the stage IIIA subgroup (86.6%) compared to the IIIB/C/D subgroup (75.8%; Fig. 2B). Meanwhile, favorable cost/insurance coverage was a more common factor for anti-PD1 therapy selection among patients in the IIIB/C/D subgroup (9.7%) compared to patients in the IIIA subgroup (2.7%; Fig. 2B).

The multivariable regression analysis identified several factors associated with adjuvant treatment choice stratified by disease substage. For patients with stage IIIA disease, SLN biopsy (OR = 23.63, 95% CI: 3.29, 169.94; P < 0.01) and CLND (OR = 5.52, 95% CI: 1.10, 27.70; P < 0.05 vs. elective dissection), and baseline LDH levels above the upper limit of normal (OR = 4.46, 95% CI: 1.06, 18.75; P < 0.05) were associated with a significantly higher probability of receiving AT. Medicaid/no insurance status (OR = 0.13, 95% CI: 0.04, 0.42; P < 0.001), ECOG PS of 2/3 (OR = 0.27, 95% CI: 0.08, 0.90; P < 0.05), and receipt of steroids (OR = 0.23, 95% CI: 0.06, 0.90; P < 0.05) were associated with a significantly lower probability of receiving AT (Supplemental Fig. 3, Additional File 1). For patients with stage IIIB/C/D disease, ECOG PS of 2/3 (OR = 0.28, 95% CI: 0.14, 0.57; P < 0.001), treatment after resection for disease recurrence (OR = 0.38, 95% CI: 0.14, 0.98; P < 0.05 vs. treatment at initial resection), and receipt of radiotherapy after resection (OR = 0.50, 95% CI: 0.27, 0.93; P < 0.05) were associated with a significantly lower probability of receiving AT (Supplemental Fig. 4, Additional File 1).

Patients in each treatment group had mean age of approximately 58 years, and the proportion of males was numerically higher among not treated patients (65.7% vs. 54.2% in the anti-PD1 subgroup, and 48.9% in the BRAF/MEK subgroup). The subgroup of not treated patients had a significantly lower proportion of commercially insured patients and higher proportion of Medicaid patients compared to the anti-PD1 and BRAF/MEK subgroups. About 17.7% of patients in the not treated group had stage IIIB/C/D disease compared with 32.0% in the anti-PD1 therapy subgroup and 40.0% in the BRAF/MEK therapy subgroup (P < 0.01; Table 1). Patients in the not treated group were significantly more likely to have higher ECOG PS (e.g., 30.4% of these patients had an ECOG PS of 2/3 compared with 10.0% in the anti-PD1 therapy subgroup and 6.7% in the BRAF/MEK therapy subgroup; P < 0.001; Table 1).

Regarding treatment utilization, the proportion of patients who received radiotherapy after resection was significantly higher among not treated patients (37.3%) compared to patients who received anti-PD1 (22.2%) or BRAF/MEK therapy (31.1%).

The multinomial logistic regression model identified several factors associated with AT choice (anti-PD1 or BRAF/MEK therapy vs. no AT; Table 3). Compared with patients who did not receive AT, stage IIIB/C/D disease was associated with a significantly higher probability of receiving treatment with anti-PD1 therapy (OR = 1.82, 95% CI: 1.09, 3.05; P < 0.05). Additionally, Medicaid/no insurance status (OR = 0.43, 95% CI: 0.22, 0.84; P < 0.05), ECOG PS of 2/3 (OR = 0.41, 95% CI: 0.22, 0.79; P < 0.01), and history of coronary artery disease/myocardial infarction (OR = 0.49, 95% CI: 0.24, 0.98; P < 0.05) were associated with a significantly lower probability of receiving anti-PD1 therapy. ECOG PS of 2/3 (OR = 0.16, 95% CI: 0.04, 0.63; P < 0.01) was also associated with a significantly lower probability of receiving adjuvant BRAF/MEK therapy.