Outcomes for pembrolizumab stratified by pemetrexed maintenance post pembrolizumab–platinum–pemetrexed induction in metastatic non-small-cell lung cancer

The introduction of immune-based therapies has introduced a new era in advanced non-small-cell lung cancer (NSCLC) treatment. Previously, platinum doublet chemotherapy was the standard first-line treatment for patients with advanced NSCLC without the then-known targetable mutations EGFR or ALK [1,2].

The use of maintenance treatments following induction therapy using agents such as gemcitabine, docetaxel and pemetrexed has been associated with significant increases in OS and/or progression-free survival (PFS) [3–5]. The PARAMOUNT randomized clinical trial enrolled patients with nonsquamous NSCLC completing four cisplatin/pemetrexed induction cycles with radiographic evidence of complete response, partial response or stable disease [6]. Enrolled patients were randomized to continuation pemetrexed + best supportive care or placebo + best supportive care. Median OS was higher for patients receiving pemetrexed maintenance (13.9 months, 95% CI: 12.8–16.0) relative to placebo (11.0 months, 95% CI: 10.0–12.5), with the corresponding hazard ratio (HR) reported as 0.78 (95% CI: 0.64–0.96) [6].

In the past few years, immune checkpoint inhibitors targeting the PD-1/PD-L1 pathway (pembrolizumab, nivolumab) have been introduced for the treatment of NSCLC [7–10]. The KEYNOTE-189 clinical trial evaluated first-line pembrolizumab + pemetrexed-platinum in patients with nonsquamous metastatic NSCLC without sensitizing EGFR/ALK aberrations, irrespective of tumor PD-L1 expression [11]. Significantly longer OS (22.0 months, 95% CI: 19.5–25.2) was observed in patients receiving pembrolizumab + pemetrexed–platinum when compared with pemetrexed-platinum (10.7 months, 95% CI: 8.7–13.6) with an HR of 0.56 (95% CI: 0.45–0.70) [12]. The pembrolizumab + pemetrexed–platinum combination regimen received expedited approval by the US Food and Drug Administration (FDA) in May 2017 and is now considered a standard of care in the first-line setting for patients with advanced or metastatic NSCLC without targetable mutations and regardless of PD-L1 expression [13].

More recently, the CheckMate 9LA trial showed that nivolumab + ipilimumab plus a limited course (two cycles) of platinum chemotherapy was associated with a significantly improved OS (HR: 0.72; 95% CI: 0.61–0.86) compared with platinum doublet chemotherapy, generating clinical interest in shorter chemotherapy regimens [14]. At the same time, real-world studies in the US using data from the Flatiron Health database observed that fewer than half of patients who initiated pembrolizumab–platinum–pemetrexed (44–48%) received five or more cycles of pemetrexed [15,16]. This is in contrast with KEYNOTE-189, where 77% of patients randomized to the pembrolizumab arm received five or more cycles of pemetrexed. In KEYNOTE-189, patients who received post-induction pembrolizumab also continued maintenance pemetrexed unless an unacceptable toxicity occurred that was attributed to only pemetrexed [11]. Another real-world study, also using the Flatiron Health database, observed significantly longer median OS (21.0 months, 95% CI: 19.3–25.2) in patients receiving maintenance treatment after induction when compared with those not receiving maintenance (9.1 months, 95% CI: 7.0–12.7); an adjusted analysis found that maintenance treatment utilization, lack of ROS1 mutation, PD-L1 tumor proportion score (TPS) ≥50%, and lower age (50–65 years) were associated with a lower risk of death, while later stage (Stage IV vs I), later diagnosis year (2018 to 2018 vs 2011 to 2017), and male sex were associated with a higher risk of death [17]. However, this study did not describe outcomes among patients who received continuation pembrolizumab by maintenance pemetrexed exposure and did not explore reasons for pemetrexed discontinuation [17].

More information is needed about the real-world administration of the KEYNOTE-189 regimen in the maintenance setting following induction. This study therefore aimed to describe maintenance treatment patterns, reasons for pemetrexed discontinuation and clinical outcomes in patients with metastatic nonsquamous NSCLC who receive induction treatment with pembrolizumab + platinum–pemetrexed in the real-world US community oncology setting.

Methods

Patient population

This was a retrospective cohort study of adult patients with metastatic NSCLC starting 1L induction therapy with pembrolizumab + platinum–pemetrexed within The US Oncology Network between 10 May 2017 and 31 August 2020. Among patients without sensitizing EGFR/ALK mutations and without or unknown ROS1 mutations, two cohorts were identified: an induction cohort, which included patients initiating induction treatment with pembrolizumab + platinum–pemetrexed (induction index visit), and a continuation pembrolizumab cohort, which included patients who received 5th to 7th cycle pembrolizumab (continuation pembrolizumab index visit) after completing four to six cycles of induction treatment without disease progression. Patients in the continuation pembrolizumab cohort were stratified into subgroups by maintenance pemetrexed exposure. Patients were required to have PD-L1 expression values, Eastern Cooperative Oncology Group performance status scores (ECOG PS) of 0 to 2 at induction index visit (± 30 days) and at least two visits excluding the induction index visit to be included in the study. Patients enrolled in clinical trials or who received treatment for another primary cancer at any time during the study observation period (10 May 2017 to 31 August 2021) were excluded. Study-eligible patients were followed longitudinally until 31 August 2021, last patient record or date of death, whichever occurred first.

Data source

Data were sourced through query of structured fields in The US Oncology Network's iKnowMed (iKM) electronic health record (EHR) system. The US Oncology Network is a network of community oncology practices with more than 1400 physicians at more than 500 locations treating more than 1.2 million patients annually [18]. iKM is an oncology specific EHR system that captures outpatient practice encounter history for patients under care, including (but not limited to) laboratory tests, diagnosis, therapy administration, line of therapy, staging, comorbidities and performance status information within The US Oncology Network. iKM captures data on outpatient medical oncology care for approximately 10% of patients newly diagnosed with cancer across the USA. Patients in the continuation pembrolizumab cohort were further examined using chart review. Additional vital status information was obtained from the Social Security Administration's Limited Access Death Master File [19].

Data analysis

Patient demographic, clinical characteristics and reasons for pemetrexed discontinuation were evaluated using descriptive statistics including mean, standard deviation, median and range for continuous variables and frequency and percentage for categorical variables. Real-world time on treatment (rwToT), rwPFS and OS were evaluated using the Kaplan-Meier method. rwToT was defined as the length of time between the first and last administration dates before discontinuation of a drug (initiation of a next line of therapy, death while receiving a drug, or a gap >120 days between last administration date and last contact date). If discontinuation criteria were not met, patients were censored at the last administration date. rwPFS was defined as the time from initiation of continuation pembrolizumab treatment to the earliest date of provider-documented progression or date of death due to any cause, censoring patients who had not died and had not progressed at the end of the study observation period by the last contact date. OS was defined as the interval between the start of continuation pembrolizumab treatment and the date of death. Patients who had not died within the study period were censored on the study end date or the last contact date available in the dataset, whichever occurred first.

We conducted an exploratory analysis after weighting the maintenance pemetrexed exposure groups by inverse probability of treatment weighting (IPTW). The propensity scores for IPTW were derived using the Toolkit for Weighting and Analysis of Nonequivalent Groups macros with generalized boosting model [20]. The propensity score calculation included age, sex, race, body mass index (BMI), smoking status, ECOG PS, time from initial diagnosis to metastatic disease, neutrophil lymphocyte ratio, albumin results, presence of brain metastases, number of other metastatic sites, documented comorbidities (hypertension, chronic obstructive pulmonary disease, renal disease, diabetes) and PD-L1 expression. Balance between the exposure groups was assessed using absolute standardized differences in standard average treatment effect (ATE) sizes pre- and post-IPTW for the selected covariates. OS, rwPFS and rwTOT were described for the balanced groups.

This study was conducted in accordance with the International Society for Pharmacoepidemiology Guidelines for Good Pharmacoepidemiology Practice [21]. The study received an exemption and waiver for informed patient consent from the US Oncology Institutional Review Board, because the study involved review of existing data and records and the information were analyzed by the study team in such a manner that research participants were not directly identified.

Results

Patient identification & characteristics

A total of 751 patients who had metastatic NSCLC without sensitizing EGFR/ALK mutations and without or unknown ROS1 mutations and started 1L induction therapy with pembrolizumab + platinum-pemetrexed between 10 May 2017 and 31 August 2020 were identified within The US Oncology Network. These were designated as the induction cohort and analyzed using structured data. Among them, 532 completed induction, 429 continued pembrolizumab and 358 underwent chart review. Among these, 241 patients qualified for the continuation pembrolizumab cohort (154 and 87 patients with and without maintenance pemetrexed, respectively) and were analyzed with supplemental chart review data (Figure 1).

Demographic and clinical characteristics of patients in the induction and continuation pembrolizumab cohorts are presented in Tables 1 & 2. In the induction cohort, median (range) age was 69 (30, 90+) years and 51.9% of patients were male. The proportion of patients ≥75 years was 26%. A total of 61% were former smokers, 19% were current smokers and 12% were never smokers. Most patients (83%) had ECOG PS scores of 0–1 while proportion of patients with PD-L1 TPS ≤1% was 40%.

In the continuation pembrolizumab cohort, among patients receiving and not receiving maintenance pemetrexed, median (range) ages were 68 (33, 85) years and 69 (52, 88) years, respectively, while proportions of males were 53 and 45%, respectively. Proportions of patients ≥75 years were 21 and 33%, respectively. For smoking status, 63 and 51%, respectively, were former smokers; 16 and 24%, respectively, were current smokers and 10 and 14%, respectively, were never smokers. Proportions of patients with ECOG PS scores of 0–1 were 71 and 76%, respectively. Proportions of patients with PD-L1 TPS ≤1% were 40 and 26%, respectively.

Treatment patterns & clinical outcomes

In the induction cohort (n = 751), 532 (71%) patients completed induction and among these patients, 429 continued pembrolizumab: 266 (50%) of them received continuation pembrolizumab with maintenance pemetrexed, and 155 (29%) received pembrolizumab alone. Only 8 (2%) received pemetrexed alone and were not further analyzed, and 103 (19%) received neither. Of all patients who initiated induction (n = 751), 266 (35%) patients received post-induction treatment with pembrolizumab and maintenance pemetrexed and 8 (1%) patients received pemetrexed alone post-induction.

Median (95% confidence interval [CI]) follow-up duration in the induction cohort from induction index visit was 10.3 (0.03, 50.1) months while median (95% CI) follow-up duration in the continuation pembrolizumab cohort from continuation index visit was 12.0 (0.5, 47.4) months. Kaplan-Meier analysis of rwToT for pembrolizumab and pemetrexed in the induction cohort is presented in Figure 2 & Supplementary Table 1. Median (95% CI) rwToT for pembrolizumab and pemetrexed were 5.1 (4.6, 5.7) months and 4.2 (3.5, 4.6) months, respectively (Figure 2).

Among patients in the continuation pembrolizumab cohort who received maintenance pemetrexed, the most frequently noted reasons for pemetrexed discontinuation were progressive disease (38%) and toxicity (29%). Among patients in the continuation pembrolizumab cohort who received pembrolizumab without maintenance pemetrexed, the most frequently noted reasons for pemetrexed being discontinued after induction were completion of planned therapy (79%) and partial response (68%) (Table 3).

Median (95% CI) rwPFS in the continuation pembrolizumab cohort evaluated using Kaplan-Meier analysis was 9.9 (8.0, 11.6) months while median (95% CI) OS was 20.3 (13.8, 26.2) months (Supplementary Table 2). Supplementary Figure 1 presents absolute standard differences in patient characteristics pre- and post-IPTW between patients who continued pembrolizumab with maintenance pemetrexed and patients who continued pembrolizumab alone while Supplementary Table 3 presents individual patient demographic and clinical characteristics pre- and post-IPTW among these groups. Median (95% CI) Kaplan-Meier OS in the IPTW groups receiving and not receiving maintenance pemetrexed were 20.3 (13.7, 27.7) months and 12.0 (9.7, 26.1) months, respectively. 6-month survival probabilities (95% CI) were 82.3% (75.0%, 87.6%) and 74.1% (61.6%, 83.1%) in the IPTW groups who received and did not receive pemetrexed, respectively, while 12-month survival probabilities were 63.4% (54.6%, 70.9%) and 49.7% (37.4%, 60.8%), respectively (Figure 3 & Table 4). Median (95% CI) Kaplan-Meier rwPFS in the IPTW groups receiving and not receiving maintenance pemetrexed were 10.3 (7.6, 13.1) months and 5.8 (3.4, 9.4) months, respectively. 6-month PFS probabilities (95% CI) were 62.4 (54.1, 69.6) months and 49.6 (38.1, 60.0) months in the IPTW groups that received and did not receive pemetrexed, respectively, while 12-month progression-free survival probabilities were 43.7 (35.5, 51.6) months and 33.9 (23.9, 44.2) months, respectively (Figure 3 & Table 4).

Discussion

The pivotal KEYNOTE-189 clinical trial led to US FDA approval of pembrolizumab in combination with pemetrexed and platinum chemotherapy for the first-line treatment of patients with metastatic nonsquamous NSCLC with no EGFR or ALK genomic tumor aberrations. This observational study aimed to describe maintenance treatment patterns, reasons for pemetrexed discontinuation and clinical outcomes in patients with previously untreated metastatic nonsquamous NSCLC who received induction treatment with pembrolizumab + platinum-pemetrexed in real-world US community oncology settings. A total of 751 patients, without EGFR or ALK mutations and with ECOG PS scores ≤2, were retrospectively identified in the iKM EHR database of The US Oncology Network during the period from 10 May 2017 to 31 August 2020. The US Oncology Network is a large and geographically dispersed health system that provides treatment for more than 1.2 million US cancer patients each year.

Among patients who initiated induction, 35% of patients received post-induction treatment with pembrolizumab and maintenance pemetrexed. These findings contrast with KEYNOTE-189, in which 77% of patients in the pembrolizumab-combination arm received maintenance pemetrexed [11]. In KEYNOTE-189, patients who continued pembrolizumab also continued maintenance pemetrexed unless an unacceptable toxicity occurred that was attributable to only pemetrexed; the observed treatment duration in KEYNOTE-189 was similar for pembrolizumab and pemetrexed (mean of 9.8 months at a median follow-up of 23.1 months) [12]. Updated data from 31.0 months of follow-up showed a median treatment duration of 7.2 months in the pembrolizumab and pemetrexed group, with 17 patients of the 410 in the pembrolizumab-pemetrexed cohort continuing to receive pemetrexed [22]. According to the 5-year update of KEYNOTE-189, seven patients of the 410 in the pembrolizumab-pemetrexed cohort were still receiving pemetrexed [23]. The pemetrexed treatment patterns observed in this study, where 52% of patients completing induction treatment received pemetrexed beyond cycle 5, are consistent with other US real-world studies reported in the literature. Velcheti et al. observed that only 48% of patients continued pemetrexed beyond cycle 4 in a study using the Flatiron Health database, and Bayo et al. reported 44% using the same data source [15,16]. Similar to The US Oncology Network, the Flatiron Health database covers a large network of community oncology practices geographically dispersed across the US. It should be noted that while our study showed median rwToT for pembrolizumab and pemetrexed in the induction cohort were 5.1 and 4.2 months, respectively, the Velcheti et al. study showed lower median rwToT for pemetrexed relative to pembrolizumab (2.8 vs 5.6 months, respectively) [16].

Consistent with Velcheti et al., our study results suggest that maintenance pemetrexed treatment patterns in the US community oncology setting deviate from those that were observed in the clinical trial setting. This observation is noteworthy, given the results from the PARAMOUNT study where maintenance pemetrexed was shown to be effective and well tolerated by patients [2]. Prior to availability of the KEYNOTE-189 pembrolizumab-based regimen, Paz-Ares et al. conducted a randomized clinical trial showing the survival benefit of maintenance pemetrexed relative to placebo in patients that did not progress during four cisplatin/pemetrexed induction cycles (HR: 0.78; 95% CI: 0.64, 0.96) [2].

Using structured data, we observed 71% of patients who initiated induction treatment with pembrolizumab + platinum-pemetrexed completed induction. Of these, 79% continued pembrolizumab treatment after four to six pembrolizumab + platinum-pemetrexed induction cycles. Unlike previous real-world studies, we incorporated a chart review component to evaluate treatment patterns and outcomes beyond induction in patients with records that were available for research and who did not have evidence of disease progression in the medical chart (n = 241). Among these patients receiving continuation pembrolizumab, 64% received continuation pembrolizumab with maintenance pemetrexed and 36% received continuation pembrolizumab alone.

Patients receiving maintenance pemetrexed with continuation pembrolizumab were younger with only 21% being 75 years or older, compared with those receiving post-induction pembrolizumab alone where 33% were 75 years or older. Also, the proportion of males was higher in the group receiving maintenance pemetrexed (53%) compared with pembrolizumab alone (45%). Differences were also noted by practice location, comorbidity, ECOG PS, number and type of metastatic sites, PD-L1 expression, and lab values. These results suggest that patient characteristics, or factors associated with these characteristics, may play a role in the use of maintenance pemetrexed in the real-world setting.

We also looked at reasons for pemetrexed discontinuation among patients who completed induction without evidence of disease progression. Among patients who never received maintenance pemetrexed in this group, the most common reasons for pemetrexed discontinuation were completion of planned treatment (79%), partial response (68%), and physician preference (9%). These reasons for discontinuation are notably different from those observed among patients stopping pemetrexed after initiating maintenance pemetrexed, where the most common reasons were progressive disease (38%), toxicity (29%) and death (7%).

Our results suggest that patient characteristics, physician choice based on patient's response during induction, and planned treatment decisions play a role in maintenance pemetrexed utilization in the community oncology setting. More information is needed to understand these factors, extension to other clinical settings, and associated impact on real-world clinical outcomes. In our study, we conducted an exploratory analysis to describe real-world clinical outcomes for maintenance pemetrexed exposure groups after using IPTW to balance the groups based on key prognostic factors. IPTW has been used in previous real-world studies to adjust for confounding factors [24]. After IPTW, KM estimates stratified by maintenance pemetrexed were consistent with improved OS and rwPFS for continuation pembrolizumab + maintenance pemetrexed relative to estimates for continuation pembrolizumab alone. These results are consistent with the known clinical benefit of pemetrexed maintenance therapy in metastatic nonsquamous NSCLC [6].

Our study findings are exploratory, and have strengths and limitations commonly associated with observational research. This retrospective observational study, which used real-world data, reflects community oncology practice trends in contrast with clinical trials that are tightly controlled. This study used data from The US Oncology Network, which is a large network of community practices in the US treating over 1.2 million patients each year. Patient chart reviews, which provide a rich source of information compared with structured data, were used to supplement data for certain variables and outcomes. This study also has several limitations, as follows. First, iKM EHR data are collected for the needs of clinical practices, and therefore data needed for research purposes may not be recorded. Therefore, data collection methods and instruments may not be consistent across practices, as with the data reporting by physicians. Second, iKM data can include coding errors of incorrect or missing data, which is a limitation of administrative databases in general. Third, the iKM EHR only contains data for patients seen by physicians and served by pharmacies within The US Oncology Network. No data will be captured for patients receiving services and procedures elsewhere, as well as drugs from pharmacies that do not serve practices in The US Oncology Network. Fourth, a patient's treatment history before any visits to practices in The US Oncology Network may not be recorded in iKM, but only in physician progress notes. Fifth, some patients coded as being treatment naive for advanced disease in the iKM EHR may have had previous chemotherapy for advanced disease in non-Network healthcare facilities. To minimize this risk, a targeted chart review was performed to verify patients' eligibility and data elements that may not be reliably captured in structured fields. With chart review, progress notes and additional free text can be reviewed to provide a much richer source of information. The induction cohort was evaluated based entirely on structured data; therefore, it includes patients that were subsequently discontinued during chart review and thus not included in the post-induction pembrolizumab cohorts. Some unmeasured confounders were not included as covariates in IPTW for exploratory analysis (e.g., response after induction treatment, time on induction treatment). Further, while IPTW was largely successful in improving balance between the groups, small differences remained and could potentially explain differences in outcomes between the two groups. Findings may not be generalizable to other community oncology settings or outside the community oncology setting.

Conclusion

Patient characteristics, physician choice based on patient's response during induction, and planned treatment decisions play a role in maintenance pemetrexed utilization in the community oncology setting. More information is needed to understand these factors, extension to other clinical settings, and associated impact on real-world clinical outcomes.