Treatment patterns and outcomes for patients with unresectable stage III and metastatic melanoma in the USA

This was an observational, retrospective, multisite and noninterventional study that collected real-world data from medical records of patients diagnosed with advanced melanoma treated at community or academic cancer centers in the USA. Patients were assigned treatment cohorts based on first-line biologic treatment: ipilimumab monotherapy, nivolumab monotherapy, nivolumab + ipilimumab, pembrolizumab monotherapy, dabrafenib + trametinib or other. The study was sponsor-blinded so that the physicians and patients remained anonymous to the source of study funding and vice versa. A chart review design was necessary to gain a deeper understanding of patients’ underlying clinical characteristics. Extracting data directly from the charts allowed for the collection of Eastern Cooperative Oncology Group (ECOG) status, BMI, mutation status, disease staging, disease progression and rationale for initiation and discontinuation of treatment. It also allowed for better identification of AEs that may not have required treatment. The target population included a random sample of patients with advanced (unresectable stage III or metastatic/stage IV) melanoma who were diagnosed prior to 1 November 2015 and received treatment for advanced melanoma from 1 January 2015 to 31 May 2016, irrespective of date of diagnosis of unresectable stage III or metastatic/stage IV melanoma. The index study period (1 January 2015 to 31 May 2016) was chosen to capture treatment patterns with outcomes observed by line of therapy (LOT), which occurred after the FDA approval of nivolumab for the treatment of advanced melanoma in 2015. This allowed for the collection of a minimum 6 months of patient follow-up data.

Clinical data were sourced from patient medical records and reflected the period from diagnosis of advanced melanoma (irrespective of unresectable or metastatic diagnosis) to the most recent visit. Month and year of last office or clinic visit were used to ensure that a representative sample of patients with advanced melanoma was enrolled within the study period.

Physicians (medical oncologists and hematologists/oncologists) were responsible for identifying eligible patients who met the inclusion criteria. From this pool of eligible patients, instructions were provided to randomly select eligible patients based on the date of initiation of index treatment within the study inclusion period (1 January 2015–31 May 2016), using random month/year generator assigned each physician/site. The process was repeated for each additional patient enrolled.

Physicians used uniform case report forms (CRFs) to extract information including patient demographics, clinical and disease characteristics, treatment patterns, treatment-related AEs and clinical outcomes.

Inclusion criteria were adults aged 18 years or older with a diagnosis of unresectable stage III or metastatic/stage IV melanoma prior to 1 November 2015, and initiation of a new LOT for advanced melanoma during the index period between 1 January 2015 and 31 May 2016, irrespective of diagnosis date. The lines of therapy were restricted to immunotherapies or targeted therapies for the treatment of advanced melanoma. Start and stop dates for each line of therapy were identified in the patient medical chart. The 1LOT was the first immunotherapy or targeted therapy given; the 2LOT was the second immunotherapy or targeted therapy – among others. ‘Other’ therapies are treatments outside of the five immunotherapies or targeted therapies of interest and are defined in results section. Complete medical history from date of diagnosis through most recent or current therapy (defined as end of data collection period) was required for medical chart extraction. Patients were excluded if they were previously enrolled in a cancer treatment-related clinical trial.

The primary outcomes included assessment of treatment patterns and clinical outcomes. Treatment patterns encompassed the year of treatment initiation, number of LOTs received, agent prescribed and rationale for treatment initiation and discontinuation. Additionally, time to initiation of each LOT (time from stage III/stage IV diagnosis to 1LOT, or time from last dose of 1LOT to first dose of 2LOT), duration of treatment (time from first dose to last recorded dose) and treatment sequence were analyzed. OS and PFS were measured from date of first recorded therapy dose during study period to date of melanoma-related death or progression, respectively. Additional outcomes included occurrence of grade III or IV AEs related to first LOT (1LOT). The proportion of patients that experienced an AE, mean number of AEs per patient and duration of AEs were assessed.

Patient demographics, clinical characteristics, treatment patterns and AE occurrence were reported using descriptive statistics; categorical statistics were described as frequency and proportion of the patient cohort, while continuous statistics were summarized as mean and standard deviation. OS and PFS were estimated descriptively using the Kaplan–Meier product limit method. Patients were censored at death, confirmed lost to follow-up or end of study, whichever occurred first. All analysis was conducted using SAS v9.4 (SAS Institute Inc., NC, USA).

This study was conducted in accordance with International Society for Pharmacoepidemiology (ISPE) Guidelines for Good Epidemiology Practices and Applicable Regulatory Requirements. The study received approval from New England Institutional Review Board on 11 January 2017.

A total of 166 physicians were screened for participation in the study, of which 45 physicians provided data. A total of 487 patients were assigned to six treatment cohorts from 45 participating ‘across all treatment groups, median follow-up time from start of 1LOT was 14.1 months (IQR: 11.0–20.3 months). Median follow-up was shortest in the nivolumab group (12.1 months [IQR: 10.6–14.5]) and longest in the dabrafenib + trametinib group (15.7 months [IQR: 11.8–19.7])’. The average age varied by 11 years between cohorts (57.1 years in nivolumab + ipilimumab vs 68.3 years in nivolumab) but the difference was not statistically significant (Table 1). A slight male preponderance was noted across cohorts (range: 50.0–66.7%), with the highest percentage in the nivolumab + ipilimumab cohort and the majority of patients were white (range: 91.6–97.1%).

Charlson Comorbidity Index scores were similar across cohorts (range: 0.4–0.9). An ECOG score of 0 was common in patients treated with nivolumab (37.7%) and nivolumab + ipilimumab (36.7%); however, most patients had an ECOG score of 1. The majority of patients in the ipilimumab, nivolumab, nivolumab + ipilimumab and dabrafenib + trametinib cohorts reported having private health insurance; whereas, most patients in the pembrolizumab cohort had government-sponsored health insurance.

The majority of patients were diagnosed at a late stage (range: 50.0–79.7%); most had stage IV at diagnosis. The most common sites for metastases were lung, lymph nodes and liver. There was one patient with a prior history of breast cancer, and one patient with prior history of prostate cancer. Less than half of patients were tested for PD-L1 expression; a third of patients treated with nivolumab and 41.1% of those treated with pembrolizumab tested positive for PD-L1 expression. The majority of patients were tested for BRAF mutations, but not all who were tested positive received a BRAF inhibitor. Of those BRAF positive, 43.5% initiated 1LOT treatment with dabrafenib + trametinib.

Of the 487 patients, 21.8% initiated 1LOT with ipilimumab, followed by 19.5% on pembrolizumab and 17.2% on dabrafenib + trametinib (Figure 1). The most common 1LOT in 2016 was nivolumab monotherapy (21.5%) (Figure 1). Of the 111 patients with BRAF mutations who initiated 2LOT, 29.7% received pembrolizumab, 25.2% nivolumab and 15.3% an ‘other’ type of treatment. The most common 1LOT and 2LOT treatments in the ‘other’ cohort were vemurafenib (n = 40), temozolomide (n = 15), dacarbazine (n = 7) and vemurafenib + cobimetinib (n = 7). The most commonly used 2LOT in 2016 was pembrolizumab monotherapy (29.2%). Of those who received 1LOT ipilimumab monotherapy, 28.3% received 2LOT, most commonly nivolumab monotherapy and pembrolizumab monotherapy (both 40.0%; Figures 2 & 3). Of patients who received 1LOT nivolumab monotherapy, 11.6% received 2LOT, most commonly an ‘other’ treatment (37.5%). Of patients receiving 1LOT nivolumab + ipilimumab treatment, only 8.3% received 2LOT, most of whom received an ‘other’ treatment (60.0%; Figures 2 & 3) . A total of 13 patients continued to 3LOT treatment. Of the 193 patients that tested BRAF positive, 68 (35.2%) moved to 2LOT, most commonly pembrolizumab (29.4%) and nivolumab (23.5%). There were 91 patients who initiated BRAF + MEK combination treatment (dabrafenib + trametinib or vemurafenib + cobemetinib), of whom nine (9.9%) progressed to a 2LOT during the study period. Gaps between the lines of therapy where noted but it is unclear if these gaps were intended as washout periods or due to other reasons.

30 patients who were initially diagnosed at an early operable stage before progressing had received systemic adjuvant therapy, with the most common types being IFN-α (63.3%) followed by pegylated IFN-α (23.3%) and ipilimumab monotherapy (6.7%). The mean time to initiation of therapy from diagnosis to first dose was shortest for dabrafenib + trametinib (3.3 weeks) and longest for pembrolizumab monotherapy (8.9 weeks). The mean duration of treatment for 1LOT therapy was shortest for patients treated with ipilimumab monotherapy (29.8 weeks) and longest for patients treated with pembrolizumab monotherapy (56.5 weeks) and nivolumab + ipilimumab (51.4 weeks) (Table 2). The duration of 2LOT therapy was shortest for ipilimumab monotherapy (8.5 weeks) and longest for dabrafenib + trametinib (35.1 weeks).

The primary reasons for 1LOT treatment choice across all therapies were improved efficacy (64.7%) and demonstrated/documented survival benefit (46.2%). The primary reason for discontinuation across 1LOT therapies was disease progression (77.3%), which was also the most common reason for discontinuation of dabrafenib + trametinib (85.7%). Drug toxicity was a commonly cited reason for discontinuation of ipilimumab monotherapy. Reasons for 2LOT discontinuation were limited to disease progression, drug toxicity and administrative reasons.

The strengths of the study included the fact that retrospective chart reviews are an efficient, reliable and verifiable method of data collection. Medical charts are often the best sources of information for documentation of cancer treatments and clinical outcomes. This study filled a gap in existing knowledge of current treatment patterns and clinical outcomes among patients diagnosed with unresectable (stage III) and/or metastatic (stage IV) melanoma who have received 1LOT immunotherapy. The stratified random selection by invitation of study physicians, planned number and geographic distribution of physicians, and robust samples of patient chart data that were collected in this study provided a rich, detailed dataset for analyses and hypothesis generation.

It should be noted that the study results are projectable within the 95% confidence level to the population of melanoma-treating oncologists and the treatment of unresectable (stage III) and metastatic (stage IV) melanoma. Another limitation is that treatment patterns represent only the practices of physicians who agreed to participate in the study and may vary from nonresponding physicians (i.e., those who refused to participate in the study and/or those who failed to complete the patient case report forms on time). An exploratory analysis was conducted, stratifying patients by the physicians practice setting. Physicians were asked to self-report their clinic setting as Academic (12), Community (11) or Academic and Community (3); 19 sites did not self report. At least 66.7%of centers reported following treatment guidelines or protocols, regardless of setting. Use of immunotherapies was similar across all practice settings. It was noted that dabrafenib + trametinib was less common in academic only settings (66.7), and ipilimumab and nivolumab monotherapies were less common in academic and community-based settings. Academic and community-based settings were also more likely to interrupt drug dosing in response to AEs.

Patient characteristics not included in the data set such as income or other family-related variables may influence the prescribing behavior but cannot be accounted for. While the potential for missing data was considered during study design and steps were taken to address these issues before study implementation, introducing multidomain preventive strategies during the pilot testing and implementation phases of this study means that missing data may still occur. For example, the death of a patient might not always be included in the patient chart.

Because of the nature of this chart review study, all clinical outcomes could be influenced by measurement error.

It was not feasible to measure the overall response rate or duration of response due to the short observation window. Future research should allow for a longer follow-up to adequately observe the duration of response.