1.Baseline characteristics of 276 patients with SCLC
In this study, the clinical information we collected included age, gender, smoking history, stage, follow-up duration, and survival status. The distribution of data from Sichuan Cancer Hospital (SCCH) and The First Affiliated Hospital of Chengdu Medical College was presented in Tables 1 and 2. The median age of patients at both centers exceeded 60 years, with the median age being 60.5 years (range:28–79 years) in SCCH and 66 years (38–81 years) in another hospital. Additionally, the proportion of smokers was higher than non-smokers at both institutions, with 58.2% and 68.8% respectively. Male patients outnumbered female patients, with males comprising 71.9% and females 28.1% in SCCH, and 71.2% male and 28.8% female in the other cohort. During the follow-up period, 66.8% of the patients died, with a median survival time of 17.55 months (range:0.3–80.8 months). At the other center, with 81.2% of the patients deceased by the end of the follow-up period, and a median survival time of 7.5 months (range:1–66 months). The staging at presentation varied between the centers. In SCCH, a greater proportion of patients were in the limited stage (LS) than the extensive stage (ES), with 72.4% in LS and 27.6% in ES. In contrast, at the other cohort, fewer patients were in LS compared to ES, with the proportions being 40% and 60%, respectively. In external independent cohort of 80 patients, the median values for TP, AST and Lymphocyte ratio were 66.75 (range:50.40–78.60), 24 (range:10.0-193.0), and 18.10 (range:3.90–44.90), respectively. The details on the laboratory indices data of 196 patients in SCCH could be found in the supplementary table 2.
Table 1
The clinical Characteristics of 196 Patients with SCLC in SCCH
Clinical characteristics | Number(%)/median(range) |
Age | 60.5 | (28–79) |
Gender | | |
Male | 141 | (71.9) |
Female | 55 | (28.1) |
Smoking history | | |
Nonsmoking | 82 | (41.8) |
Smoking | 114 | (58.2) |
Event | | |
Alive | 65 | (33.2) |
Dead | 131 | (66.8) |
Stage | | |
I-III | 142 | (72.4) |
IV | 54 | (27.6) |
Time(month) | 17.55 | (0.3–80.8) |
Table 2
The clinical Characteristics of 80 Patients with SCLC in external independent validation cohort
Clinical characteristics | Number(%)/median(range) |
Age | 66.5 | (38–81) |
Gender | | |
Male | 57 | (71.2) |
Female | 23 | (28.8) |
Smoking history | | |
Nonsmoking | 25 | (31.2) |
Smoking | 55 | (68.8) |
Event | | |
Alive | 15 | (18.8) |
Dead | 65 | (81.2) |
Stage | | |
I-III | 32 | (40.0) |
IV | 48 | (60.0) |
Time(month) | 7.5 | (1–66) |
TP(g/L) | 66.75 | (50.40–78.60) |
Lym ratio (%) | 18.10 | (3.90–44.90) |
AST(U/L) | 24 | (10.0-193.0) |
2.Selection of 6 predictive indicators
We conducted prognostic analysis on 61 laboratory indicators, including 23 routine blood indicators, 22 clinical biochemical indicators, 7 coagulation function indicators, and 9 infectious disease indicators for 196 small cell lung cancer patients in SCCH, and 21 markers were found to be significant (Fig. 1A). Subsequently, lasso-cox regression was used to select the best predictive factors, and among them, 16 candidates had the minimum lambda value (Fig. 1B and Table 3). As shown in Table 3, separate Cox regression analyses were performed for each of the 16 prognostic indicators for survival. We obtained 8 indicators with p-values < 0.01, among which 6 indicators with HR < 1 were considered as protective factors for SCLC, namely ALB (HR = 0.51, 95%CI: 0.36–0.72, p < 0.001), total protein (TP) (HR = 0.51, 95%CI: 0.36–0.72, p < 0.001), Lymphocyte ratio (HR = 0.54, 95%CI: 0.398–0.77, p < 0.001), RBC (HR = 0.56, 95%CI: 0.4–0.8, p = 0.0012), Ret low F intensity (HR = 0.59, 95%CI: 042-0.84, p = 0.0031), TCH (HR = 0.61, 95%CI: 0.43–0.86, p = 0.0047), while the other two indicators with HR > 1 were AST (HR = 1.6, 95%CI: 1.1–2.3, p = 0.0077), Ret high F intensity (HR = 2.0, 95%CI: 1.3–3.1, p = 0.0018), respectively. Subsequent multivariate analysis revealed that TP (HR = 0.6, 95%CI: 0.4–0.89, p = 0.012), Lymphocyte ratio (HR = 0.67, 95%CI: 0.46–0.96, p = 0.028), TCH (HR = 0.68, 95%CI: 0.48–0.96, p = 0.03), and AST (HR = 1.7, 95%CI: 1.2–2.4, p = 0.0049) were independent risk factors for OS. Further Cox regression was applied to filter these four markers. The indices AST (p < 0.001), Lymphocyte ratio (p < 0.001), TP (p = 0.005) had p-values less than 0.01, and therefore were selected for subsequent modeling (Fig. 1C). We also performed prognostic analyses on four clinical variables: gender, age, smoking history, and stage. Ultimately, age (HR = 1.7, 95%CI: 1.2–2.4, p = 0.0047), smoking history (HR = 1.6, 95%CI: 1.1–2.4, p = 0.0076), and stage (HR = 1.7, 95%CI: 1.2–2.4, p = 0.0056) were identified as independent prognostic factors (Fig. 1D).
Table 3
Univariate and Multivariate Cox proportional hazards Regression Analysis of OS
Variable | Univariate Multivariate |
HR (95%CI) p value HR (95%CI) p value |
ALB | 0.51(0.36–0.72) | < 0.001 | 0.78(0.52–1.2) | 0.24 |
TP | 0.51(0.36–0.72) | < 0.001 | 0.6(0.4–0.89) | 0.012 |
Lym ratio | 0.54(0.38–0.77) | < 0.001 | 0.67(0.46–0.96) | 0.028 |
RBC | 0.56(0.4–0.8) | 0.0012 | 0.81(0.55–1.2) | 0.28 |
Ret Low F Intensity | 0.59(0.42–0.84) | 0.0031 | 0.74(0.5–1.1) | 0.13 |
TCH | 0.61(0.43–0.86) | 0.0047 | 0.68(0.48–0.96) | 0.03 |
MG | 0.63(0.45–0.9) | 0.01 | | |
PreS1 antigen of HBV | 0.64(0.45–0.9) | 0.011 | | |
HCV Ab | 0.66(0.45–0.6) | 0.028 | | |
HBeAg | 0.68(0.46–0.99) | 0.044 | | |
MPV | 0.7(0.5–0.99) | 0.046 | | |
MCV | 1.4(1.0–2.0) | 0.047 | | |
AST | 1.6(1.1–2.3) | 0.0077 | 1.7(1.2–2.4) | 0.0049 |
D-D | 1.5(1.1–2.1) | 0.021 | | |
Neu ratio | 1.5(1.1–2.1) | 0.022 | | |
Ret high F Intensity | 2(1.3–3.1) | 0.0018 | 1.3(0.76–2.1) | 0.37 |
3.Construction of the Nomogram with six indicators
Scores were assigned to these six independent prognostic factors based on the cox regression and this prognostic model was created in discover cohorts. Elevated levels of AST, staging at the ES, older age, and a history of smoking were associated with a higher probability of poor prognosis. SCLC patients with AST levels higher than 25 U/L scored 66, while those with lower levels scored 89. Smoking history scored 95, compared to 66 for non-smokers. Those in extensive stage scored 66, while those in the limited stage scored 42. Patients older than 58.5 years scored 66, whereas those younger scored 0. Conversely, high levels of blood total protein (TP) and the Lymphocyte Ratio are beneficial for patient survival. Patients with TP content above 70.4 g/L scored 66, while those with lower levels scored 100. Lymphocyte ratio higher than 22.9% scored 66, and those with lower levels scored 89 (Fig. 2A). According to this score criteria, we converted the laboratory data clinical pathological information of all 196 patients into nomogram scores and reorganized them based on the score levels (Fig. 2B). As shown in the figure, the majority of patients with high nomogram scores was older, and had elevated AST levels, a more advanced stage, lower levels of Lymphocyte ratio and TP, a history of smoking. In addition, patients with dead events were mainly concentrated in the high-score segments, and they also had shorter survival time, which indicated that SCLC patients with different ages, stages, and laboratory data could be distinguished partly by nomogram score. Independent 80 cohort performed the nomogram score transformation as well. Patients with different clinical characteristics and outcomes were also discriminated (Supplementary Fig. 1B).
4.Validation of nomogram model
For two-year survival rates, the predicted values in the discover group were close to the actual values, while in the validation group, the prediction for patients with higher survival rates was very accurate, but the fit was relatively lower for patients with worse prognosis. For three-year survival rates, the training group's predictions for patients with worse prognosis were more accurate than those for patients with better prognosis, while in the validation group, the predictions are more accurate for both patients with good and poor prognosis, but the fit with the predicted values is not high for the middle portion (Fig. 2C), but overall, it was observed that the cox predicted probability was generally consistent with the actual value. Besides, time-dependent receiver operating characteristic (time-ROC) was utilized to evaluate the predictive accuracy of nomogram. In the discover and validation cohort, the AUC for 2-year and 3-year OS were 0.74, 0.74 and 0.64, 0.74, respectively (Fig. 2D). As for external independent center, the AUC for 2-year and 3-year were even better, with 0.92 and 0.82 respectively (Supplementary Fig. 1A).
5.Comparison of predictive accuracy between novel model and previous prognostic system
The cutoff value of the discover group was determined based on the scoring system, which was used for dividing 196 SCC patients and independent 80 cohort into high risk and low risk groups. The novel predictive model allowed for more accurate differentiation among patients with varying survival periods. The prognosis model based on stage identified 54 individuals in ES and 142 in LS. The number of survivors at 3 years was 14 for ES and 56 for LS. The 6-year survival rates were 3.7% for ES and 7% for LS, with a P-value of 0.0084. In contrast, stratification through the nomogram score delineated groups into high-risk (63 individuals) and low-risk (133 individuals). At three years, the survival figures stood at 5 and 65 individuals for high-risk and low-risk groups, respectively, whereas the five survival numbers were 0 and 12 individuals, respectively, with a smaller P-value. The new model has thus made corrections to survival rates across different periods, exhibiting statistically more significant implications (Fig. 3A-B). Observations from another center also mirrored this trend (Fig. 3C-D), which indicated that the nomogram improved survival predictions for SCLC patients.
6.Novel model distinguished risk stratification for SCLC patients with the same stage.
According to the nomo score system, we categorized SCLC patients in ES into high-risk and low-risk groups and conducted an analysis of overall survival periods. The OS for three-year in high-risk group was 7.4%, and the six-year OS was 0, which was significantly lower than the 44.4% and 7.4% for the low-risk group, with a p-value of 0.00032 (Fig. 4A). As for patients in LS, The three-year and six-year survival rates for the high-risk group were 8.3% and 0%, respectively, compared to 50% and 9.4% for the low-risk group, with a p-value less than 0.0001 (Fig. 4B). In independent cohort, ES patients had poorer survival rates, with the longest survival time being 31 months. After grouping according to the nomo score, we were able to distinguish patients in ES with different outcomes (p = 0.0021). 34 individuals were categorized into the low-risk group, of whom two survived over 24 months, while none of the fourteen high-risk patients survived beyond 24 months (Fig. 4C). Patients in LS exhibited the same trend, with the overall survival rate of the low-risk group being higher than that of the high-risk group (p = 0.032). In the low-risk group, five people survived over two years and one for more than four years, while in the high-risk group, only one person survived for 24 months (Fig. 4D).