Real-World Treatment Patterns and Clinical Outcomes in Patients With Extensive-Stage Small Cell Lung Cancer Treated With First-Line Platinum-Based Chemotherapy and ≥ 2 Subsequent Lines of Therapy in the United States.

Key Summary Points

Introduction
Lung cancer is the most common malignancy and the leading cause of cancer-related mortality worldwide, accounting for 12.4% of cancer diagnoses and 18.7% of cancer deaths in 2022 [1]. Approximately 15% of all lung cancer cases are small cell lung cancer (SCLC), which is a high-grade pulmonary neuroendocrine tumor with an aggressive clinical course [2]. Consistent with its rapidly progressive nature, approximately two-thirds of patients with SCLC are diagnosed with extensive-stage SCLC (ES-SCLC), as defined by the presence of distant metastases and/or tumors extending beyond an area that can be treated within a radiation field [2]. Prognosis for patients with ES-SCLC is poor, with a 5-year survival rate of < 4% [3].
The standard first-line (1L) treatment for ES-SCLC for many years was platinum-based chemotherapy (PBC), comprising cisplatin or carboplatin in combination with the topoisomerase II inhibitor etoposide [2, 4]. More recently, immune checkpoint inhibitors (ICIs) have been incorporated into standard-of-care 1L therapy [5], with the anti-programmed death ligand 1 (PD-L1) agents atezolizumab and durvalumab approved for use in combination with 1L PBC [6], followed by maintenance monotherapy based on the randomized phase 3 IMpower133 and CASPIAN studies, respectively [7, 8]. However, despite many patients initially responding to these 1L regimens, responses are generally not durable, and the disease relapses in most patients within 6 months [9–11]. Studies in patients with ES-SCLC treated with 1L therapy outside of clinical trials have shown that median overall survival (OS) is 7.1–8.0 months in those who did not receive a PD-L1 inhibitor and 8.3–13.5 months in those who did [10–12].
Patients with ES-SCLC who progress on or after 1L PBC-containing therapy have limited available treatments. Until recently, the only approved options in the United States (US) were the non-platinum chemotherapeutic agents lurbinectedin (a DNA-binding alkylating agent) and oral/intravenous topotecan (a topoisomerase I inhibitor) [2], which are associated with median OS of 9.3 and 6.0–8.1 months, respectively, as second-line (2L) treatments for SCLC in phase 2/3 clinical trials [13–16]. Patients with a chemotherapy-free interval (CTFI) ≥ 3 months after 1L therapy may also be candidates for PBC rechallenge [5], which was associated with a median OS of 7.5 months in a phase 3 trial in this patient subgroup [17]. Tarlatamab, a delta-like ligand 3-targeted bispecific T-cell engager, received accelerated US approval in May 2024 for the treatment of patients with ES-SCLC with progression on or after PBC [18], and subsequently demonstrated significantly prolonged OS compared with standard-of-care chemotherapy as 2L treatment in the phase 3 DeLLphi-304 study [19]. However, due to the risk of cytokine release syndrome and neurologic toxicity, tarlatamab administration requires step-up dosing and complex patient monitoring [20, 21]. There remains, therefore, an unmet need for effective new therapies with acceptable safety profiles, and simple administration and monitoring requirements to treat relapsed ES-SCLC.
Real-world studies elucidating patient characteristics, treatment patterns, and clinical outcomes across successive treatment lines can play a valuable role in informing the clinical development of novel therapies for ES-SCLC. However, few such studies are currently available, and data following the shift in treatment landscape caused by the integration of ICIs into the 1L setting are limited [11, 12, 22, 23]. We conducted a retrospective study to evaluate patient characteristics and treatment sequencing from 1L to fourth line (4L), and clinical outcomes with third-line (3L) therapy, overall and in clinically relevant subgroups, among a large, real-world, US nationwide cohort of patients with ES-SCLC who received 1L PBC between January 2018 and June 2023.

Methods

Study Design and Data Source
This retrospective, observational study used data from the US nationwide Flatiron Health electronic health record (EHR)-derived database, which collects longitudinal, patient-level, de-identified, real-world data from patients receiving care at community (~ 75%) and academic (~ 25%) cancer centers across the US [24]. The Flatiron Health network includes access to over 5 million patient records from over 280 oncology practices (community and academic) at more than 800 unique sites of care in the US. The study also utilized Flatiron Health Spotlight data, in which real-world response data were manually abstracted from EHRs to complement the main dataset. As this was a retrospective study using de-identified patient data from an EHR-derived database, it was exempt from informed consent procedures and Institutional Review Board approval. Permission was obtained to access and use data from Flatiron Health.

Study Period and Patient Population
The study period was selected as January 1, 2018, through December 31, 2023, to ensure that the dataset reflected recent changes in the treatment landscape following the approvals of ICIs and lurbinectedin [25]. To allow for at least 6 months of potential follow-up, the patient identification period was from January 1, 2018, through June 30, 2023. Patients were eligible for inclusion in the study if they were ≥ 18 years of age at ES-SCLC diagnosis, had a histopathologically confirmed diagnosis of SCLC, and had ≥ 2 documented clinical visits on or after the start of the initial study period, with no prior diagnosis of non-small cell lung cancer. Patients must also have had extensive-stage disease at initial diagnosis (abstracted from clinical documents as SCLC-specific stage by the Flatiron algorithm) and initiated 1L PBC (defined as any treatment regimen containing cisplatin or carboplatin) during the patient identification period. Patients with limited-stage disease at initial diagnosis were excluded. Patients were also excluded if they received clinical trial study drugs (as determined by the oncologist-defined, rule-based Flatiron line of therapy algorithm) in any line of treatment, or oxaliplatin in 1L, 2L, 3L, or 4L.
The overall population was defined as the 1L cohort. Patients who initiated ≥ 1, ≥ 2, and ≥ 3 subsequent lines of therapy (as defined by the Flatiron line of therapy algorithm) before the end of the patient identification period were included in the 2L, 3L, and 4L cohorts, respectively. Within each cohort, the index date was the date of initiation of the corresponding line of therapy. The follow-up period spanned 1 day after the treatment index date to the last confirmed structured activity or treatment activity before the end of the study period.

Clinical Outcomes
The primary objective was to describe demographic and clinical characteristics, and treatment patterns and sequencing in the 1L, 2L, 3L, and 4L cohorts. Study variables included age, sex, race, Eastern Cooperative Oncology Group performance status (ECOG PS) at the index date, smoking history, care setting, the presence of brain metastases at the index date (identified by International Classification of Diseases [ICD], Ninth and Tenth Revision codes in structured electronic medical record data only), and CTFI (defined as the time interval from the end of 1L PBC to the start of any 2L therapy). For the description of treatment patterns, treatments were classified as follows: PBC + topoisomerase inhibitor; PBC + anti-PD-(L)1 + topoisomerase inhibitor; PBC “other”; anti-PD-(L)1 monotherapy; anti-PD-(L)1 “other”; any lurbinectedin-containing regimen; topoisomerase inhibitor(s) only; chemotherapy “other”; or other. A detailed description of the criteria for treatment categorization is presented in Table S1.
Secondary objectives were to describe real-world overall survival (rwOS), real-world time to treatment discontinuation or death (rwTTD/D), real-world time to next treatment or death (rwTTNT/D), and confirmed real-world response rate (rwRR) in the 3L cohort. rwOS was defined as the time interval from the 3L index date to the date of death from any cause. rwTTD/D was defined as the time interval from the 3L index date until the date of treatment discontinuation (defined by the date of 4L treatment initiation, or the last contact date before a period of ≥ 120 days without structured activity but with subsequent confirmed structured activity) or death from any cause. rwTTNT/D was defined as the time interval from the 3L index date until the date of 4L treatment initiation or death from any cause. The confirmed rwRR was defined as the proportion of patients who achieved a best overall response of complete or partial response based on all real-world response assessments from the 3L index date until the earliest occurrence of real-world disease progression, initiation of 4L therapy, or end of follow-up. Confirmed rwRR was assessed in patients with ≥ 2 real-world response assessments ≥ 28 days apart after the 3L index date.
Subgroup analyses were conducted to assess 3L and 4L treatment patterns, as well as clinical outcomes from 3L therapy initiation according to ECOG PS, treatment in prior lines (i.e., anti-PD-[L]1 therapy and PBC rechallenge), and CTFI, which is associated with greater chemoresponsiveness [26]. Given that clinical guidelines recommend PBC rechallenge for patients with longer CTFI [5], it has the potential to impact both subsequent treatment choices and clinical outcomes in later treatment lines. Treatment patterns were analyzed in patients who had received 1L anti-PD-(L)1 therapy (defined as 1L treatment with nivolumab, pembrolizumab, cemiplimab, durvalumab, or atezolizumab, in combination with PBC), patients with ECOG PS 0–1 at the cohort index date, and patients who had received PBC rechallenge (defined as 2L PBC). Clinical outcomes in the 3L cohort were analyzed for the same subgroups, as well as for patients with an ECOG PS of 2 at the 3L index date, or CTFI < 90, ≥ 90, < 180, or ≥ 180 days. Individual patients could be included in multiple subgroups.

Statistics
All analyses were descriptive. Continuous variables were summarized as medians with ranges, and categorical variables as frequencies with percentages. Treatment patterns were visualized in Sankey plots. rwOS, rwTTD/D, and rwTTNT/D were analyzed using Kaplan–Meier methodology. Patients without an event before the end of follow-up were censored at the last activity date within the database. For each time-to-event outcome, medians and rates at 6 months and 1, 2, and 3 years were estimated, along with 95% confidence intervals (CIs). For confirmed rwRR, the 95% CI was calculated using the Clopper–Pearson method. Statistical analyses were performed using SAS® version 9.4 (SAS Institute Inc., Cary, NC, USA). No formal statistical comparisons were performed.

Results

Patient Population
A total of 2666 adults with an initial diagnosis of ES-SCLC initiated 1L PBC during the patient identification period (Fig. 1). After exclusion of 93 patients due to receipt of clinical trial drugs during the study period (n = 90) or oxaliplatin in 1L–4L treatment (n = 3), the final 1L cohort comprised 2573 patients. Of the patients in the 1L cohort, 992 (38.6%), 344 (13.4%), and 114 (4.4%), respectively, were included in the 2L, 3L, and 4L cohorts. Therefore, 38.6% of patients in the 1L cohort, 34.7% in the 2L cohort, and 33.1% in the 3L cohort went on to receive at least one subsequent line of therapy during the patient identification period, demonstrating substantial attrition across lines of therapy among this patient population.
In the 1L cohort, median age at the index date was 68.0 years (range 30.0–85.0) (Table 1). The proportion of female (n = 1300; 50.5%) and male (n = 1273; 49.5%) patients was approximately equal, a majority were white (n = 1826; 71.0%), and almost all had a history of smoking (n = 2526; 98.2%). ECOG PS was 0 or 1 in 1451 patients (56.4%), while brain metastases were recorded on or before the 1L index date for 306 patients (11.9%).

Demographic and clinical characteristics were generally similar across the 1L, 2L, 3L, and 4L cohorts. However, the proportion of patients with brain metastases increased with each successive treatment line, at 11.9%, 23.2%, 29.9%, and 35.1%, respectively. In addition, the percentage of patients with ECOG PS 0–1 was higher in the 2L, 3L, and 4L cohorts than in the 1L cohort. Furthermore, patients with CTFI ≥ 180 days accounted for a greater proportion of patients in the 4L cohort compared with the 2L and 3L cohorts.

Treatment Patterns
In the 1L cohort, PBC was administered with a topoisomerase inhibitor in combination with an anti-PD-(L)1 agent in approximately two-thirds of patients (n = 1711; 66.5%), and without an anti-PD-(L)1 agent in approximately one-third of patients (n = 837; 32.5%) (Fig. 2; Table 2). Thereafter, there was no clear standard of care in the 2L, 3L, or 4L cohorts. In the 2L cohort, lurbinectedin-containing regimens (monotherapy or combinations) were the most common treatment (n = 263; 26.5%). A total of 176 patients (17.7%) in the 2L cohort received PBC in combination with both anti-PD-(L)1 therapy and a topoisomerase inhibitor, 96 patients (9.7%) received PBC in combination with a topoisomerase inhibitor alone, and 19 patients (1.9%) received “other” PBC regimens. Thus, across these three treatment categories, 291 patients (29.3%) received further PBC in 2L.
In the 3L cohort, the most common treatments were lurbinectedin-containing regimens and topoisomerase inhibitors (administered as monotherapy or in combination only with another topoisomerase inhibitor), each received by 75 patients (21.8%). Other regimens administered in the 3L cohort included other chemotherapy (n = 54; 15.7%), anti-PD-(L)1 monotherapy (n = 45; 13.1%), and PBC + topoisomerase inhibitor (n = 35; 10.2%). In the 4L cohort, the most common treatments were topoisomerase inhibitor(s) only (n = 34, 29.8%), other chemotherapy (n = 27, 23.7%), and lurbinectedin-containing regimens (n = 21, 18.4%). In both 2L and 3L, nivolumab + ipilimumab was the most common regimen in the category of anti-PD-(L)1 “other” (2L, n = 68 [6.9%]; 3L, n = 13 [3.8%]).
In subgroup analyses, 3L and 4L treatment patterns in patients with 1L anti-PD-(L)1 therapy (3L, n = 191; 4L, n = 59), ECOG PS 0–1 (3L, n = 225; 4L, n = 72), or PBC rechallenge (3L, n = 125; 4L, n = 50) generally reflected those seen in the overall cohorts (Table S2). The use of anti-PD-(L)1 monotherapy in 3L or 4L was less common in patients who had received 1L anti-PD-(L)1 therapy than in other subgroups. Compared with patients in other subgroups, patients who were rechallenged with PBC were more likely to receive 3L lurbinectedin-containing regimens or 4L topoisomerase inhibitor(s) only and less likely to receive other chemotherapy in 3L.

Clinical Outcomes in the 3L Cohort
Median rwOS from the 3L index date was 4.53 months (95% CI, 3.71–5.39) (Fig. 3A). rwOS rates from the 3L index date decreased from 40.9% (95% CI, 35.6–46.1) at 6 months to 20.6% (95% CI, 16.2–25.3) at 1 year, 3.6% (95% CI, 1.5–7.0) at 2 years, and 1.8% (95% CI, 0.4–5.2) at 3 years. Median rwTTD/D from the 3L index date was 2.56 months (95% CI, 2.27–2.79) (Fig. 3B). rwTTD/D rates from the 3L index date were 14.6% (95% CI, 11.0–18.7) at 6 months, 5.0% (95% CI, 2.9–7.9) at 1 year, and were not estimable at 2 or 3 years. Median rwTTNT/D from the 3L index date was 2.92 months (95% CI, 2.69–3.12), with TTNT/D rates at 6 months and 1 year of 21.1% (95% CI, 16.9–25.7) and 6.5% (95% CI, 4.1–9.6), respectively, while 2- and 3-year rwTTNT/D rates were not estimable (Fig. 3C). Among 77 patients in the 3L cohort who were evaluable for response, confirmed rwRR was 11.7% (95% CI, 5.5–21.0).
In subgroup analyses of the 3L cohort, patients with ECOG PS 0–1 at 3L index had a median rwOS of 5.85 months (95% CI, 4.90–6.37), median rwTTD/D of 2.79 months (95% CI, 2.40–3.09), and median rwTTNT/D of 3.22 months (95% CI, 2.89–3.84), indicating broadly comparable clinical outcomes relative to the overall 3L cohort, with a numeric trend for longer rwOS (Table 3). In contrast, patients with ECOG PS 2 had numerically worse clinical outcomes, especially regarding rwOS, although the subgroup size was limited (n = 68; Table S3). Clinical outcomes in patients with 1L anti-PD-(L)1 therapy or PBC rechallenge were mostly similar to those observed in the overall 3L cohort (Table 3). In patients with PBC rechallenge, there was a trend for numerically longer median rwOS (6.80 months; 95% CI, 4.53–8.02) relative to the overall 3L cohort, whereas median rwTTD/D (2.79 months; 95% CI, 2.33–3.35) and median rwTTNT/D (3.52 months; 95% CI, 2.83–3.98) were similar. In patients with 1L anti-PD-(L)1 therapy, median rwOS was 4.90 (95% CI, 3.22–5.85), median rwTTD/D was 2.63 (95% CI, 2.30–3.02), and median rwTTNT/D was 3.02 months (95% CI, 2.69–3.71).

In subgroup analyses that dichotomized the 3L cohort into those with CTFI < 90 days (n = 149) and ≥ 90 days (n = 195), median rwOS was 3.84 months (95% CI, 2.99–4.83) and 5.59 months (95% CI, 3.81–6.64), median rwTTD/D was 2.56 months (95% CI 2.20–2.89) and 2.56 months (95% CI, 2.14–2.99), and median TTNT/D was 2.89 months (95% CI, 2.43–3.12) and 3.02 months (95% CI, 2.69–3.65), respectively (Table S3). The trend for prolonged rwOS in patients with longer CTFI was even more pronounced among patients with CTFI ≥ 180 days (n = 80); however, other efficacy outcomes in this group and among those with CTFI < 180 days (n = 264) were similar to those observed when using a 90-day cutoff.

Discussion
In this large retrospective study of patients with ES-SCLC in the US treated with 1L PBC, we found that, beyond 1L therapy, treatment patterns were heterogeneous, response rates were low, and survival was limited. Furthermore, attrition across lines of therapy was substantial. Together, these results underscore the lack of standard of care in later lines and highlight the substantial unmet need for new treatments for patients with relapsed ES-SCLC.
Treatment patterns and sequencing were broadly similar to previous reports using real-world data from patients with ES-SCLC in the US [12, 23]. Consistent with guidelines, 1L PBC was mostly used in combination with anti-PD-(L)1 agents and a topoisomerase inhibitor [5]; however, a third of patients received 1L PBC without anti-PD-(L)1 agents, which may reflect approvals for these agents partway through the study period [6] and/or physician reluctance to use anti-PD-(L)1 therapy in some patients (e.g., due to autoimmune disease) [27].
Beyond 1L, treatment patterns were fragmented, with no clear standard, in line with the wide range of clinical guideline-recommended options [5, 28] and limited head-to-head clinical trial data in this setting. Lurbinectedin was the most common 2L regimen, but was used less frequently than expected, possibly due to the timing of its approval [29] or uncertainty following the failure of the confirmatory ATLANTIS trial [30]. PBC rechallenge was relatively frequent, especially in 2L, as expected given clinical guideline recommendations that such treatment can be considered in patients with CTFI 3–6 months and is preferred if the CTFI is longer [5]. The use of anti-PD-(L)1 agents in combination with PBC is associated with prolonged CTFI, which may increase the proportion of patients who would be candidates for PBC rechallenge [9, 12]. The relatively frequent use of PBC rechallenge with anti-PD-(L)1 agents was surprising, as these combinations are not specifically recommended in guidelines [5, 28]. However, this result may reflect how line-of-therapy information was captured, since regimen-based algorithms in real-world databases are derived from drug prescription or administration dates, and gaps in treatment information may trigger a new line of therapy [31].
There was substantial heterogeneity in 3L treatment. Approximately a quarter of patients received PD-(L)1 inhibitors (roughly equally split between monotherapy and combinations), and lurbinectedin-based regimens or topoisomerase inhibitors were each used by just over a fifth of patients. Similar heterogeneity in 3L treatments was observed in two other Flatiron Health database studies capturing real-world data that spanned the approval of ICIs for 1L treatment of ES-SCLC [12, 23]. In one study, which included 790 patients with ES-SCLC receiving 3L therapy after any 1L therapy (2013–2022), the most common 3L treatments were ICI-containing regimens (excluding anti-PD-L1 in combination with PBC and etoposide) in 2018 (40%) and 2019 (39%), with reduced use after 2019 (≤ 19%), accompanied by an increase in lurbinectedin and topotecan use [12]. In the other study, which included 326 patients with SCLC (one-third with limited-stage disease at diagnosis) who received 1L PBC (2018–2021), treatment patterns were broadly similar to those reported here. However, single-agent ICI use was higher (in just over a quarter of patients) [23]. Greater ICI use during 2018–2021 was expected based on the approval and subsequent withdrawal of 3L indications for nivolumab and pembrolizumab in ES-SCLC [25].
Among the 344 patients who received 3L treatment in this study, median rwOS from 3L therapy initiation was only 4.53 months, which is similar to the 4.1 months reported among patients with ES-SCLC in one of the aforementioned Flatiron Health database studies [12]. By contrast, longer median rwOS from 3L therapy initiation (5.3 months) was reported among patients with SCLC who received 1L PBC in the other Flatiron Health database study [23], which may reflect the inclusion of patients with limited-stage disease, who have a better prognosis [2]. However, a similarly longer median rwOS from 3L therapy initiation of 5.8 months was reported in a retrospective US electronic medical record-derived study of 113 patients with ES-SCLC (2012–2022) [32]. Although patient characteristics were similar to those in the present study, the median age was lower (58 years vs. 66.5 years) and the sample size was smaller, which may, in part, explain the observed differences in survival. Additional real-world studies within and outside the US that predated the approvals of ICIs for 1L treatment of ES-SCLC have shown consistently poor median rwOS (2.7–5.8 months) [33–38].
This study provides some of the first estimates in this setting for rwTTD/D and rwTTNT/D as proxies for progression-free survival (PFS) not requiring manual abstraction [39]. The median rwTTD/D of 2.56 months and median rwTTNT/D of 2.92 months are comparable with prior real-world PFS estimates of 2.3–3.1 months [23, 32–34]. Furthermore, the confirmed rwRR of 11.7% was within the range reported in previous studies (7.7–21.3%) [23, 32–34, 38].
Clinical outcomes in subgroups of interest were similarly poor compared with the overall 3L cohort. rwTTD/D (median 2.63–2.79 months) and rwTTNT/D (median 3.02–3.52 months) showed no notable differences across the subgroups, nor between the subgroups and the overall 3L cohort (median rwTTD/D, 2.56 months; median rwTTNT/D, 2.92 months). Small numerical increases in median rwOS relative to the overall 3L population (4.53 months) were observed in the ECOG PS 0–1 (5.85 months) and PBC rechallenge subgroups (6.80 months); although not formally tested, these differences are unlikely to be statistically significant based on the overlapping 95% CIs.
Importantly, clinical outcomes among patients with ECOG PS 0–1 (the most commonly enrolled performance status in clinical trials) closely mirrored the broader 3L cohort, supporting the external validity and real-world applicability of clinical trial data. Poor clinical outcomes in patients with ECOG PS 2 in this study are consistent with the established negative prognostic impact of worse performance status [40]. The numerically improved rwOS among patients with 2L PBC rechallenge may be attributable to its greater use in patients with better health status [41]. In addition, PBC rechallenge is recommended for patients with longer CTFI [5], which is associated with greater chemoresponsiveness in subsequent lines [42], consistent with trends for longer rwOS in patients with CTFI ≥ 90 or ≥ 180 days in the present study. The poor clinical outcomes after 3L therapy in patients who received 1L anti-PD-(L)1 agents are in line with prior studies including patients with ES-SCLC [12, 23]. Overall, the poor clinical outcomes despite PBC rechallenge, longer CTFI, or 1L anti-PD-(L)1 use support the need for new 3L therapies, irrespective of treatment and response in prior lines.
Brain metastases were recorded at 1L index in 11.9% of patients, rising sequentially across lines of therapy to 35.1% in the 4L cohort, which was lower than the expected frequency of ~ 50–60% in later treatment lines [2]. Our study is likely to have underestimated the prevalence of brain metastases due to its reliance on ICD coding, which can be incomplete and hence a limitation of EHR data-based studies [43]. Given the high incidence of brain metastases in patients with SCLC, the potential for them to develop is a key factor in clinical decision-making. The likely underreporting of brain metastases in this study is an important gap that will need to be filled by future studies. Nonetheless, the data highlight the substantial unmet need for treatments with intracranial activity.
This analysis leveraged a large and nationally representative ES-SCLC dataset, which provided comprehensive data on patient characteristics, treatments, and clinical outcomes, including mortality and longitudinal follow-up, to offer timely insights into evolving treatment patterns. The mortality variable within the database is validated against the National Death Index [44], and the inclusion of several years of follow-up after the adoption of chemoimmunotherapy as a 1L standard of care allowed for the capture of data reflective of the recent treatment paradigm for ES-SCLC. However, our study did not capture data on the use of tarlatamab, which received accelerated approval in the US after the study period [20]; this would make a valuable addition to future studies. Other elements not captured by our study, owing to the nature of the database used, included whether patients had received radiotherapy after 1L treatment, prognostic or predictive factors that may have affected treatment patterns and outcomes, as well as reasons for treatment discontinuation and quality-of-life data. Further limitations include the potential for survivorship bias due to the exclusion of patients not treated with 1L PBC, and selection bias in confirmed rwRR analyses due to the requirement for ≥ 2 assessments ≥ 28 days apart, which may not correspond to real-world assessment schedules. Interpretation may be limited by relatively small sample sizes in some 3L cohort subgroups and in the 4L cohort and its subgroups. Furthermore, generalizability beyond the US population may be limited. This study also shares the inherent limitations of EHR-derived datasets, including incomplete coding, potential misclassification, and underreporting. Possible approaches to mitigate against some of these limitations in future studies include using different databases or registries, capturing data from different geographic locations, and conducting studies prospectively.

Conclusion
This analysis of a large, nationwide, multi-institutional US database demonstrates that patients with ES-SCLC treated with 1L PBC-containing therapy had heterogeneous treatment patterns in subsequent lines, with no clear standard of care. Patients receiving 3L therapy had universally poor clinical outcomes, irrespective of ECOG PS, prior treatment, and CTFI, underscoring an urgent need for novel treatments in this setting.

Supplementary Information
Below is the link to the electronic supplementary material.