A simplified scoring system for predicting treatment response in limited-stage small-cell lung cancer (EAST score)

ABSTRACT

Aims

This study aimed at developing a scoring system (EAST score) to predict recurrence after chemoradiotherapy in limited-stage small-cell lung cancer (LS-SCLC).

Patients & Methods

Treatment-naïve LS-SCLC patients receiving concurrent chemoradiotherapy (CCRT) (N = 234) or sequential chemoradiotherapy (N = 53) were retrospectively reviewed. Using data from CCRT population, clinical and radiological variables associated with disease progression were identified. Selected variables were assigned numerical scores based on their estimated hazard ratios (HRs), and the EAST score was established.

Results

EAST score incorporated N3 disease and serum biomarkers (lactate dehydrogenase, pro-gastrin-releasing peptide, and cytokeratin-19 fragment). In the CCRT population, progression-free survival (PFS) was significantly shorter in the high-risk group (EAST score ≥ 2) than the low-risk group (EAST score ≤ 1) (median, 9.4 months vs. 20.6 months; HR [95% confidence interval (CI)], 2.09 [1.50–2.91]). As for the model performance, the 1- and 2-year area under the curve values for PFS were 0.68 and 0.65, respectively. Overall survival was also shorter in the high-risk group (HR [95% CI], 1.49 [1.02–2.16]). Similar trends were observed in the sequential chemoradiotherapy population (HR for PFS [95% CI], 2.43 [1.07–5.53]).

Conclusions

EAST score effectively predicts recurrence risk in LS-SCLC, demonstrating the necessity for developing new treatment strategies for high-risk patients.

Plain Language Summary

What is this article about? This study introduced the EAST score, a new system to predict recurrence risk in people with limited-stage small-cell lung cancer (LS-SCLC) receiving chemoradiotherapy. Chemoradiotherapy is the main treatment for LS-SCLC, but physicians do not have a good way to guess how well it will work for each patient. Using information from LS-SCLC patients, the researchers found certain biomarkers that are linked to the cancer recurrence. The EAST score was created using these biomarkers.

What were the results? Out of 1,496 patients with SCLC, 234 with LS-SCLC had treatment with concurrent chemoradiotherapy, and 53 had sequential chemoradiotherapy. The EAST score included N3 disease and some blood test results (lactate dehydrogenase, pro-gastrin-releasing peptide, and cytokeratin-19 fragment). In patients who had concurrent chemoradiotherapy, those with a high-risk EAST score (≥2 points) had short progression-free survival (9.4 months on average) compared to the low-risk patients (≤1 point, 20.6 months). High-risk patients also had short overall survival (34.2 months) compared to low-risk patients (53.0 months on average). Similar results were seen in patients who had sequential chemoradiotherapy.

What do the results of the study mean? This study shows that the EAST score is helpful for physicians to predict cancer outcomes using results of routine tests. Patients with a high-risk score might need stronger or different treatments to help them live longer.

1. Introduction

Small-cell lung cancer (SCLC) is a highly aggressive malignancy that is categorized as limited-stage (LS) or extended-stage (ES) for treatment decision-making [1,2]. Concurrent chemoradiotherapy (CCRT), which involves administration of cisplatin and etoposide concurrently with accelerated hyperfractionated thoracic radiotherapy (AHF-TRT), is the standard treatment for LS-SCLC [3]. In elderly patients or those with poor Eastern Cooperative Oncology Group performance status (ECOG PS), sequential chemoradiotherapy (SCRT) is an alternative [4]. Prophylactic cranial irradiation (PCI) is recommended in patients who respond to chemoradiotherapy [5]. However, only a third of LS-SCLC patients survive 5 years and the median progression-free survival (PFS) is approximately 1 year [6–8]. This survival data indicates that many patients experience early disease progression after initial treatment and that the efficacy of chemoradiotherapy for LS-SCLC is heterogeneous.

Although several prognostic models have been developed to predict SCLC outcomes, they are suboptimal for LS-SCLC patients who undergo chemoradiotherapy because of numerous variables related to treatment modalities and site of distant metastases [9,10]. Studies reporting prognostic factors in LS-SCLC patients have noted that locally advanced tumors (stage III), N3 disease, and inadequate response to systemic therapy are associated with high risk of relapse [11–14]. However, these studies did not combine these risk factors nor provide precise criteria for predicting chemoradiotherapy response. Moreover, although several laboratory findings, including tumor markers, are associated with SCLC, their role in predicting treatment response or disease progression remains unclear [15,16].

This study aimed to identify clinical and radiological factors that predict treatment response to chemoradiotherapy by analyzing data from LS-SCLC patients who underwent CCRT. A machine learning-based algorithm was employed to develop a clinical scoring system, the EAST score, which is both objective and user-friendly. The clinical utility of this score was evaluated through internal validation. In addition, we investigated the applicability of the EAST score in a separate dataset consisting of LS-SCLC patients who underwent SCRT.

2. Material & methods

2.1. Patients and study design

We retrospectively reviewed consecutive LS-SCLC patients who received chemoradiotherapy in our hospital (National Cancer Center Hospital East) from April 2009 to June 2022. The in-hospital registry contained data regarding patient age, gender, date of registration, tumor histology, location of primary lesion, tumor stage, ECOG PS, and baseline laboratory data. Patients aged ≥18 years with histologically confirmed LS-SCLC and ECOG PS score 0–2 who underwent four cycles of platinum-based chemotherapy and received either 45 Gy TRT in 30 twice-daily fractions or 50 Gy TRT in 25 once-daily fractions were eligible for study inclusion. LS-SCLC was defined as disease in which the primary tumor and any associated lymph node involvement is limited to a single radiation field [17]. We included patients who had discontinued TRT because of toxicity but were still able to complete the planned dose [6]. Patients with multiple cancers or those already enrolled in clinical trials were excluded. Patients who received TRT concurrently with any cycle of chemotherapy were assigned to the CCRT population; those who received it after completing four cycles of chemotherapy were assigned to the SCRT population.

2.2. Data collection and candidate variables

The following parameters were extracted as candidate variables for disease progression: age; gender; ECOG PS score (2 vs. 0–1); smoking cessation; serum levels of albumin, lactate dehydrogenase (LDH), C-reactive protein, carcinoembryonic antigen, cytokeratin-19 fragment (CYFRA), neuron-specific enolase, pro-gastrin-releasing peptide (ProGRP), and hemoglobin; monocyte count; monocyte-to-lymphocyte ratio; neutrophil-to-lymphocyte ratio; and platelet-to-lymphocyte ratio [16,18–22]. All patients were staged using the eighth edition TNM classification for lung cancer [23]. T status (T1/T2/T3/T4) and N status (N2/N3) were selected as candidate variables.

Next, we gathered data regarding chemotherapy, TRT type and schedule, TRT parameters such as mean lung dose (MLD) and percentage of total lung volume receiving 5 Gy/20 Gy or more (V5/V20), as well as other therapeutic modalities (e.g., surgery or PCI) and recurrence. The following parameters were selected as candidate variables: MLD, V5, V20, timing of TRT, and presence of ipsilateral pleural effusion on computed tomography [24,25]. Early TRT was defined as initiation of TRT with the first or second cycle of chemotherapy; late TRT was defined as initiation with the third or fourth cycle.

2.3. Statistical analyses

PFS was defined as the time from treatment initiation to disease progression or death from any cause. Overall survival (OS) was defined as the time from treatment initiation to death from any cause. Because chemoradiotherapy is the definitive treatment for LS-SCLC, we used PFS as the endpoint for treatment efficacy in the process of variable selection. In the subsequent analyses, both PFS and OS were used as endpoints.

2.3.1. Development of the EAST score

Data from patients who underwent CCRT were analyzed to develop the EAST score. A random survival forest (RSF) was used to select candidate variables that were significantly associated with shorter PFS. Next, we performed multivariable Cox proportional hazards analysis to estimate hazard ratios (HRs) of the selected variables. Missing data were imputed via multiple imputation using chained equations. Twenty datasets were imputed and the results were combined using Rubin’s rules [26]. Serum ProGRP level was treated as a categorical variable with three categories, where the lower cutoff was set at 100 pg/mL and the upper cutoff was determined based on the log-rank maximization method [27]. Other selected parameters were treated as categorical variables consisting of two categories. The upper limit of normal range was determined as the cutoff for laboratory data, and this cutoff value was verified by the log-rank maximization method [28]. The logarithms of the point estimates of HRs were compared, then rounded to the nearest integers to assign scores to each variable. The sum of the scores was defined as the patient’s EAST score. As a sensitivity analysis, we treated the serum ProGRP level as a continuous variable and performed multivariate analysis alongside other explanatory variables.

2.3.2. Internal validation of the EAST score

To evaluate the efficacy of the scoring system, we calculated the EAST score for each patient. Patients with missing data in the variables used for the EAST score were excluded. The CCRT population was stratified into low- and high-risk of disease progression using the median EAST score as the cutoff value. Survival curves for PFS and OS were generated using the Kaplan – Meier method and compared using the log-rank test. Patients were further stratified into subgroups based on individual EAST scores and separate Kaplan – Meier curves were generated for PFS and OS for each subgroup. The discriminability of the scoring system was evaluated using time-dependent receiver operating characteristic (ROC) curves and Harrel’s C statistics. Time-dependent ROC curves provided area under the curve (AUC) values for 1-year and 2-year predictions.

The model was calibrated using two methods: (1) the same dataset that was used to develop the EAST score and (2) resampling. In the resampling process, a bootstrap sample, which matched the size of the original dataset, was created through repeated random sampling. Cross-validation was conducted using data points that were not included in the bootstrap sample. This entire process was iterated 10,000 times to evaluate the model’s internal validity. Time-dependent calibration plots were then generated for 1-year, 2-year, and 3-year PFS.

2.3.3. Sensitivity analyses regarding OS in the CCRT population

To assess the association between OS and each variable of the EAST score, we conducted a multivariable Cox proportional hazards analysis for OS in the CCRT population. Missing values were addressed through multiple imputation, as outlined in 2.3.1.

In this dataset, OS data included a proportion of censored cases. We conducted two additional sensitivity analyses for OS between high- and low- risk groups in the CCRT population:

1. Based on the inverse probability censoring weighting (IPCW) method, IPCW-adjusted Kaplan – Meier curve was created to estimate the median OS. The adjusted HR and 95% confidence interval (CI) were estimated using a robust variance estimator.

2. In order to quantitatively assess whether the estimated OS HR was robust against censoring cases, a tipping point analysis was performed, which was a generalization of the best-worst case imputation method. Using δ as a multiplicative constant, the OS data of censored cases in each group were multiplied by δ, and the HR was estimated. This process was iterated across continuous values of δ to identify the threshold at which the HR lost statistical significance [29,30].

2.3.4. Evaluation of the applicability of the EAST score

Using data from patients who underwent SCRT as an independent dataset, we evaluated the applicability of the EAST score. Patients in the SCRT population were categorized into low- and high-risk groups using the same cutoff value as in the CCRT population. The log-rank test was used to compare PFS between groups.

The current study was performed according to the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) [31]. Statistical analyses were performed using R version 4.2.2. RSF analysis was performed using the open source “RandomForestSRC” package, and the process of calibration was performed using the “pec” package.

3. Results

3.1. Patient characteristics in the CCRT group

The study flowchart is shown in Supplementary figure S1. Among 1496 patients with treatment-naïve SCLC, 234 patients with LS disease underwent CCRT as first-line treatment. Patient characteristics are shown in Table 1. Median age was 67 years and 173 patients (74%) were male. ECOG PS score was 0 or 1 in 229 (98%). N status was N2 in 124 patients (53%) and N3 in 73 (31%). Two hundred seven patients (88%) had stage III disease. The median serum ProGRP level was 345 pg/mL. Monocyte-to-lymphocyte ratio was high (>0.25) in 93 patients (40%) [10]. Neutrophil-to-lymphocyte ratio was high (>4) in 35 patients (15%) [18]. Platelet-to-lymphocyte ratio was high (>200) in 183 patients (78%) [20]. Regarding TRT, V20 ≤ 35% and MLD ≤20 Gy were achieved in most patients [1]. Median follow-up was 27.0 months. Median PFS and OS in the CCRT population were 13.3 months and 40.1 months, respectively (Figure 1).

Table 1.

Patient characteristics in the concurrent chemoradiotherapy group.

Variable	Overall (N = 234)
Age (y)	67 (61–72)
Male gender, n (%)	173 (74)
Performance status, n (%)
0	129 (55)
1	100 (43)
2	5 (2)
Smoking cessation, n (%)	118 (50)
T status, n (%)
T1	112 (48)
T2	56 (24)
T3	28 (12)
T4	38 (16)
N status, n (%)
N0-N1	37 (16)
N2	124 (53)
N3	73 (31)
Stage, n (%)
Stage II	27 (12)
Stage III	207 (88)
Pleural effusion, n (%)	35 (15)
Laboratory data
Albumin [g/dL]	4.2 (4.0–4.4)
LDH [U/L]	208 (177–245)
CRP [mg/dL]	0.18 (0.06–0.46)
CEA [ng/mL]	3.6 (2.1–6.6)
CYFRA [ng/mL]	2.4 (1.8–3.2)
NSE [ng/mL]	26 (16–50)
ProGRP [pg/mL]	345 (79–960)
Hemoglobin [g/dL]	13.9 (12.8–14.9)
Monocyte count [102/µL]	4.0 (3.1–4.9)
LMR <4, n (%)	93 (40)
NLR >4, n (%)	35 (15)
PLR >200, n (%)	183 (78)
Parameters of TRT
MLD [Gy]	11.0 (9.4–13.2)
V5 [%]	38.9 (31.6–47.8)
V20 [%]	22.5 (18.6–26.9)

Data are presented as numbers (%) or medians (interquartile range).

LDH, lactate dehydrogenase; CRP, C-reactive protein; CEA, carcinoembryonic antigen; CYFRA, cytokeratin-19 fragment; NSE, neuron-specific enolase; ProGRP, pro-gastrin-releasing peptide; MLR, monocyte-to-lymphocyte ratio; NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; TRT, thoracic radiotherapy; MLD, mean lung dose; V5/V20, percentage of the total lung volume receiving 5 Gy/20 Gy or more.

Figure 1.

Kaplan – Meier curves of (a) Progression-free survival and (b) Overall survival in the concurrent chemoradiotherapy population.

3.2. Development of the EAST score

Figure 2 depicts the RSF results, where N3 disease, serum LDH, ProGRP, and CYFRA levels were of high importance and selected as predictive variables. Supplementary figure S2 shows the partial dependence plots of variables with importance > 0.005. The upper cutoff of serum ProGRP level was set at 1000 pg/mL (Supplementary figure S3). In multivariable Cox proportional hazards analysis, each variable received a score based on the log HR, and the clinical scoring system was developed (Table 2). Very high ProGRP level (>1000 pg/mL) showed the largest HR, followed by high ProGRP level (100–1000 pg/mL), elevated CYFRA (>3.5 ng/mL), elevated LDH (>240 U/L), and N3 disease. The cutoff for LDH calculated by the log-rank maximization method was 235 U/L, and this cutoff value was deemed reasonable.

Figure 2.

Result of random survival forest. Candidate variables and variable importance for each are shown.

PS2, Performance status score 2; LDH, lactate dehydrogenase; CRP, C-reactive protein; CEA, carcinoembryonic antigen; CYFRA, cytokeratin 19 fragment; NSE, neuron-specific enolase; ProGRP, pro-gastrin-releasing peptide; MLR, monocyte-to-lymphocyte ratio; NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; MLD, mean lung dose; V5/V20, percentage of total lung volume receiving 5 Gy/20 Gy or more; CCRT, concurrent chemoradiotherapy.

Table 2.

Multivariable survival analysis for progression-free survival and the clinical scoring system (EAST score).

	adjusted HR [95% CI]	P value	Score
N3 disease	1.37 [0.97–1.93]	0.080	1
LDH >240 U/L	1.46 [1.01–2.11]	0.045	1
ProGRP
>1000 pg/mL	2.09 [1.32–3.33]	0.002	2
100–1000 pg/mL	1.68 [1.12–2.52]	0.012	1
CYFRA >3.5 ng/mL	1.64 [1.10–2.46]	0.017	1

HR, hazard ratio; CI, confidence interval; LDH, lactate dehydrogenase; ProGRP, pro-gastrin-releasing peptide; CYFRA, cytokeratin-19 fragment.

In the multivariate analysis treating ProGRP as a continuous variable, the hazard of ProGRP increased monotonically with its rise. For every 100 pg/mL increase in ProGRP, the hazard ratio increased by 1.02 (95% CI: 1.01–1.03) (Supplementary table S1), indicating the validity of dividing serum ProGRP levels into several categories.

3.3. Stratification of the CCRT group by EAST score

Supplementary figure S4 shows the distribution of the EAST score (median score, 2). Ten patients had missing values and were excluded. Patients with an EAST score ≥ 2 were assigned to the high-risk group (N = 117). Patients with score ≤ 1 were assigned to the low-risk group (N = 107). Table 3 provides an overview of patient characteristics in each group. AHF-TRT was performed in 88% of patients in the high-risk group and 86% in the low-risk group. Response to CCRT was achieved in more than 95% and subsequent PCI was administered in 64% in both groups. Brain metastases occurred in 16% of high-risk patients and 21% of low-risk patients. Extrathoracic recurrence other than brain occurred in 35% and 20% of patients, respectively. PFS was significantly shorter in the high-risk group than the low-risk group (median, 9.4 months vs. 20.6 months; HR, 2.09; 95% CI, 1.50–2.91; Figure 3(a)). OS was also significantly shorter in the high-risk group (median, 34.2 months vs. 53.0 months; HR, 1.49; 95% CI, 1.02–2.16; Figure 3(b)).

Table 3.

Patient characteristics in the concurrent chemoradiotherapy group according to risk subgroup.

Variable	High-risk (N = 117)	Low-risk (N = 107)
Age (y)	67 (61–72)	68 (62–72)
Male gender, n (%)	89 (76)	79 (74)
Performance status, n (%)
0	49 (42)	76 (71)
1	63 (54)	31 (29)
2	5 (4)	0 (0)
N3 disease, n (%)	66 (56)	4 (4)
Stage, n (%)
Stage II	1 (1)	25 (23)
Stage III	116 (99)	82 (77)
LDH >240 U/L, n (%)	55 (24)	6 (3)
ProGRP (pg/mL)	853 (223–1874)	141 (56–392)
>1000 pg/mL, n (%)	52 (44)	0 (0)
100–1000 pg/mL, n (%)	44 (38)	60 (56)
CYFRA >3.5 ng/mL, n (%)	44 (19)	5 (2)
Chemotherapy, n (%)
Platinum/etoposide	116 (99)	105 (98)
Other platinum-based regimen	1 (1)	2 (2)
AHF-TRT, n (%)	103 (88)	92 (86)
Late CCRT*, n (%)	42 (36)	18 (17)
Treatment response, n (%)
CR	33 (28)	43 (40)
PR	79 (68)	62 (58)
SD	1 (1)	0 (0)
PD	4 (3)	2 (2)
PCI, n (%)	75 (64)	68 (64)
Site of recurrence, n (%)
Inside the irradiation field	27 (23)	21 (20)
Thorax†	40 (34)	24 (22)
Brain	19 (16)	23 (21)
Other sites outside the thorax	41 (35)	21 (20)

*Beginning TRT concurrently with the third or fourth cycle of chemotherapy.

^†Recurrence within the thoracic cavity and outside the irradiated field, e.g., contralateral lung field or contralateral hilar lymph node.

LDH, lactate dehydrogenase; ProGRP, pro-gastrin-releasing peptide; CYFRA, cytokeratin-19 fragment; AHF-TRT, accelerated hyperfractionated thoracic radiotherapy; CCRT, concurrent chemoradiotherapy; CR, complete response; PR, partial response; SD, stable disease; PD, progressive disease; PCI, prophylactic cranial irradiation.

Figure 3.

Kaplan – Meier curves of (a) Progression-free survival and (b) Overall survival in the low- and high-risk groups of the concurrent chemoradiotherapy population.

HR, hazard ratio; CI, confidence interval.

Kaplan – Meier curves for PFS and OS stratified according to EAST score demonstrated a discernible trend of poorer PFS and OS with increasing EAST score (Figure 4).

Figure 4.

Kaplan – Meier curves of (a) Progression-free survival and (b) Overall survival in patients stratified according to EAST score in the concurrent chemoradiotherapy population.

mPFS, median progression-free survival; mOS, median overall survival; NR, not reached.

3.4. Evaluation of the model performance

According to the time-dependent ROC curves, the 1- and 2-year AUCs were 0.68 and 0.65, respectively (Supplementary figure S5). The Harrel’s C statistic value was 0.623. The EAST score demonstrated good calibration in both the dataset used for model development and in the internal validation using bootstrap samples. In patients with very high risk of recurrence, the EAST score tended to provide estimates of recurrence risk that were slightly higher than the actual observed risk (Supplementary figure S6).

3.5. Sensitivity analyses regarding OS

The result of the multivariable analysis for OS using variables of the EAST score is shown in Supplementary table S2. The point estimate of the HR for each variable suggested that all variables were independently associated with OS.

OS was censored in 58/107 patients (54%) in the low-risk group and 53/117 patients (45%) in the high-risk group. From the IPCW-adjusted Kaplan-Meier curve, the estimated median OS was 53.0 months for the low-risk group and 34.2 months for the high-risk group. The adjusted OS HR was 1.51 (95% CI: 1.03–2.22; Supplementary figure S7). In the tipping point analysis, even when the OS of the high-risk group was hypothetically overestimated, the point estimate of HR exceeded 1.4 in many cases. (Supplementary figure S8A). The tipping points at which the statistical significance disappeared were constantly above the line of equal scaling between the two groups (Supplementary figure S8B). These sensitivity analyses would support the robustness of the EAST score for the stratification of OS.

3.6. Stratification of the SCRT group by EAST score

Supplementary table S3 shows the characteristics of 53 LS-SCLC patients who underwent SCRT and the groups stratified by EAST score. Median follow-up was 17.1 months. Median PFS and OS were 9.1 months (Supplementary figure S9A) and 18.7 months, respectively. Excluding one patient with a missing value, PFS was significantly shorter in the high-risk group than the low-risk group (N = 14 and 11, respectively; median, 8.3 months vs. 23.4 months; HR, 2.43; 95% CI, 1.07–5.53; Supplementary figure S9B).

4. Discussion

In this study, we identified variables that predict treatment response to CCRT in patients with treatment-naïve LS-SCLC and developed an easy-to-use clinical scoring system, the EAST score, to categorize them according to risk of recurrence. Patients in the high-risk group had significantly shorter PFS and OS than the low-risk group. The EAST score was also effective for risk stratification in patients who underwent SCRT. This is the first study to introduce criteria for predicting treatment response by analyzing data exclusively from LS-SCLC patients undergoing chemoradiotherapy.

In LS-SCLC patients undergoing CCRT, the Kaplan – Meier curve for PFS revealed a substantial hazard within the first year, indicating a high incidence of disease recurrence. The curve’s slope gradually declined thereafter, implying a lower risk of recurrence. Additionally, some patients managed to remain disease-free for up to five years, which has also been reported in previous studies [7,8]. These findings suggest that LS-SCLC patients are a heterogeneous population with varying responses to CCRT. Nonetheless, to date, there has been no attempt to stratify LS-SCLC patients based on recurrence risk. In a previous study, stage III disease was associated with a higher risk of recurrence than stage I – II. However, approximately 85% of LS-SCLC patients in the study had locally advanced tumors and using only clinical staging for risk stratification is insufficient [11].

A RSF identified that N3 disease and elevated serum levels of LDH, ProGRP, and CYFRA were variables significantly associated with poor response to CCRT. To select candidate variables, we extracted parameters that were previously reported to be predictive factors or have been included in prognostic nomograms [16,18–22]. A machine learning-based algorithm was used for variable selection and multivariable analysis showed that all these variables were independent predictors of recurrence. Previous SCLC studies did not identify CYFRA as a prognostic determinant. However, elevated serum CYFRA level is observed in some patients with SCLC [16]. Evaluation of multiple factors in the RSF has unveiled the significance of CYFRA as an independent predictor of recurrence. Importantly, these variables are routinely measured or evaluated in patients with lung cancer, and therefore, the EAST score can be easily calculated.

PCI reportedly improves prognosis in LS-SCLC [5]. However, since it is selectively administered to patients who respond to chemoradiotherapy, PCI is an intermediate factor in predicting treatment response. Consequently, we did not include it as a covariate in the multivariable analysis. The decision to perform PCI in this study was at the treating physician’s discretion and considered factors such as ECOG PS and comorbidities. Notably, PCI rate was well balanced between the high- and low-risk groups (64% in both groups of the CCRT population). Similarly, early CCRT is associated with a better prognosis than late CCRT [24]. However, in cases where the irradiation field is extensive, late CCRT is preferred to maintain MLD or V20 within acceptable limits. In the high-risk group of CCRT patients, late CCRT was more frequently selected because these patients frequently had larger tumors or higher number of lymph node metastases. Late CCRT was not selected as a predictive variable in the RSF, suggesting that its effect was appropriately surrogated by other factors.

In addition to LS-SCLC patients who underwent CCRT, we evaluated the applicability of the EAST score in those who underwent SCRT. These analyses indicated that the EAST score effectively predicts treatment response to not only CCRT but also SCRT. Meanwhile, in patients with ES-SCLC treated with chemotherapy alone, prognostic factors such as ECOG PS score, age, and number and location of metastases have already been identified, which differ from the variables included in the EAST score [9,10,32]. Furthermore, treatment with chemotherapy is not curative in ES-SCLC. Therefore, it seems challenging to directly apply the EAST score to predict prognosis in ES-SCLC patients.

In the risk classification based on the EAST score, patients with a score of ≥ 2 were classified as high-risk, and those with a score of 0–1 were classified as low-risk, to ensure that the number of patients in each group was approximately equal. The median PFS for patients with an EAST score of 2 was 10.2 months, which is shorter than the average PFS of approximately 12 months for LS-SCLC, supporting the classification of those with a score of ≥ 2 as high-risk. The median PFS of patients in the high-risk group was only 9.4 months in the CCRT population, indicating poor response to chemoradiotherapy. Currently, novel therapies for ES-SCLC, such as antibody-drug conjugates and bispecific T-cell engagers, are under development [33,34]. The combination of these agents with TRT may potentially improve the prognosis of LS-SCLC, especially among patients with high risk of recurrence. The EAST score can help to identify these patients. Additionally, to date, stratification factors have not been sufficiently identified to include when designing randomized controlled trials (RCTs) involving LS-SCLC patients. Including this predictive score as a stratification factor in RCTs could facilitate more precise assessment and comparison of treatment efficacy between the study and control groups.

Several limitations of this study should be acknowledged. First, it is essential to confirm the external validity of this scoring system, especially in patients receiving CCRT. The upper cutoff value of ProGRP was established as 1000 pg/mL in this study, which needs to be confirmed. Considering that half of the patients in the low-risk group experienced recurrence within the first 2 years, it is imperative to determine the optimal cutoff value for risk stratification. We are currently planning to validate the EAST score by using data from RCTs previously conducted in Japan for patients with LS-SCLC (JCOG0202 and JCOG1011). The results will soon be present. Second, the EAST score is a risk classification system based solely on clinical factors, and its performance could potentially be improved by incorporating genomic prognostic factors. For instance, mutations in TP53 and the absence of RB1 mutations have been suggested as indicators of poor prognosis [35]. Integrating these genomic insights with clinical parameters may enable more accurate predictions of patient outcomes. In addition, it remains unclear whether the EAST score alone is sufficient as a stratification factor for OS. In this study, multiple sensitivity analyses demonstrated that OS can be stratified using the EAST score. Since age and ECOG PS are generally recognized as prognostic factors in advanced cancers, incorporating these variables could potentially improve the accuracy of predicting outcomes after tumor recurrence [36]. Finally, the EAST score for each patient may have variations owing to several factors. For example, baseline laboratory data can be influenced by differences in measurement assays, blood collection techniques, and the presence of comorbidities, such as impaired renal function. Moreover, a time delay between the diagnosis and the start of treatment may lead to an underestimation of the EAST score.

5. Conclusions

In conclusion, we have identified clinical factors that predict treatment response in LS-SCLC patients and have established a novel scoring system based on N status in combination with pre-treatment laboratory findings. Patients with a high-risk of recurrence require new treatment strategies to improve their outcome.

Funding Statement

This paper was not funded.

Article highlights

• This study established a clinical scoring system (EAST score) to predict response to chemoradiotherapy in limited-stage small-cell lung cancer (LS-SCLC) patients.

• Using machine learning-based approach, N3 stage, serum levels of lactate dehydrogenase, pro-gastrin-releasing peptide, and cytokeratin-19 fragment were selected as prognostic variables among clinical and radiological parameters.

• The EAST score is calculated from these variables, and it effectively stratified LS-SCLC patients who underwent concurrent chemoradiotherapy (CCRT) into high and low recurrence risk groups. The high-risk group showed significantly shorter progression-free survival and overall survival compared to the low-risk group.

• The EAST score was also applicable to patients who underwent sequential chemoradiotherapy (SCRT).

• Based on the EAST score, patients with a high-risk of recurrence require new treatment strategies to improve their outcome. Furthermore, this scoring system enables clinical trial designs that consider the heterogeneity in treatment response among LS-SCLC patients.

Author contributions

Yu Ito: Conceptualization, Methodology, Investigation, Writing – Original Draft. Yoshitaka Zenke: Conceptualization, Methodology, Writing – Original Draft. Tetsuya Sakai: Writing – Review & Editing. Yuji Shibata: Writing – Review & Editing. Hiroki Izumi: Writing – Review & Editing. Kaname Nosaki: Writing – Review & Editing. Shigeki Umemura: Writing – Review & Editing. Shingo Matsumoto: Writing – Review & Editing. Kiyotaka Yoh: Writing – Review & Editing. Masaki Nakamura: Investigation, Writing – Review & Editing. Hidehiro Hojo: Investigation, Writing – Review & Editing. Takehiro Izumo: Writing – Review & Editing. Koichi Goto: Writing – Review & Editing, Supervision.

Disclosure statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Dr. Ito reports receiving personal fees from Chugai Pharmaceuticals; Dr. Zenke reports receiving personal fees from Amgen, AstraZeneca, Boehringer Ingelheim, Bristol-Meyers Squibb, Chugai Pharmaceuticals, Kyowa Kirin, Lilly, MSD, Nihon Kayaku, Novartis, Ono, Pfizer, Taiho, and Takeda; Dr. Sakai reports receiving personal fees from AstraZeneca, Chugai Pharmaceuticals, Merck Serono, MSD, Novartis, Takeda, and Thermo Fisher Scientific; Dr. Shibata reports receiving personal fees from AstraZeneca, Bristol-Meyers Squibb, Chugai Pharmaceuticals, Lilly, Merck, Ono, Pfizer, and Takeda; Dr. Izumi reports receiving personal fees from Bristol-Meyers Squibb, Chugai Pharmaceuticals/Roche, Merck, MSD, Ono, and Takeda; Dr. Nosaki reports receiving personal fees from AstraZeneca, Chugai Pharmaceuticals, Janssen, Lilly, Merck, MSD, Novartis, Ono, Pfizer, Taiho, and Takeda; Dr. Umemura reports receiving personal fees from Chugai Pharmaceuticals; Dr. Yoh reports personal fees from AbbVie, Amgen, AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Chugai Pharmaceuticals, Daiichi Sankyo, Janssen, Kyowa Kirin, Lilly, Merck, Novartis, Ono, Otsuka, Taiho, and Takeda; Dr. Nakamura reports receiving personal fees from AstraZeneca, Illumina, Varian; Dr. Izumo reports receiving personal fees from AstraZeneca, Boehringer Ingelheim, Chugai Pharmaceuticals, Lilly and MSD; Dr. Goto reports receiving personal fees from Amgen, Amoy Diagnostics, AstraZeneca, Bayer HealthCare Pharmaceuticals, Bayer Yakuhin, Boehringer Ingelheim, Blueprint, Daiichi Sankyo, Eisai, Guardant Health, Haihe Biopharma, Janssen, Lilly, Medpace, Merck, Nippon Kayaku, Novartis, Ono, Otsuka, Riken Genesis, Taiho, and Takeda. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.

Ethical declaration

This study was approved by the ethics committee of National Cancer Center Hospital East (approval number, 2020–136). The requirement for written informed consent was waived owing to the retrospective nature of the study. Instead, the patients were provided with an opportunity to opt out. All methods were performed in accordance with relevant guidelines and regulations.

Registry and the Registration No. of the study/trial: 2020–136.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/14796694.2024.2444858