Estimating the Cure Proportion in Stage IA Lung Adenocarcinoma: A population-based study

doi:10.21203/rs.3.rs-3146056/v1

Download PDF

Research Article

Estimating the Cure Proportion in Stage IA Lung Adenocarcinoma: A population-based study

https://doi.org/10.21203/rs.3.rs-3146056/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 31 Oct, 2023

Read the published version in BMC Pulmonary Medicine →

You are reading this latest preprint version

Objectives

Using a mixed cure model, this study aimed to investigate factors influencing the cure, recurrence, and metastasis of stage IA lung adenocarcinoma.

Materials and Methods

A total of 1,064 patients who underwent video-assisted thoracoscopic pneumonectomy were included. Variable screening was performed using the random forest algorithm and least absolute shrinkage and selection operator. The mixed cure model was used to identify factors affecting patient cure and survival, and a sequential analysis was performed on 5%, 10%, and 20% of the presentational subtype simultaneously. The receiver operating characteristic was used to explore the best model and to construct a nomogram for predicting cure rate.

Results

The median follow-up was 58 (range: 3–115) months. Results from the cure part of the mixed model indicated that the predominant subtype, presentational subtype, and tumor diameter were prognostic factors affecting the patients' cure rate. Based on these factors, a nomogram was constructed to predict the cure rate. The survival part indicated that the predominant subtype was the only factor influencing patients' recurrence and metastasis. A sequential analysis of the presentational subtype showed no significant effect on survival (P > 0.05). Regardless of the recording mode, no significant improvement was observed in the model's discriminative ability. Only a few postoperative pathological specimens showed lymphovascular invasion (LVI); however, the survival curve suggested a significant effect on patient survival.

Conclusion

After excluding the existence of long-term survivors, the predominant subtype was the only factor influencing patient recurrence and metastasis. Although LVI is rare in stage IA, its significance for patient prognosis cannot be ignored.

mixed cure model

lung adenocarcinoma

stage IA

histological subtype

prognostic analysis

Primary lung cancer is currently one of the most common cancers in the world, with lung adenocarcinoma (LUAD) being the main type of lung cancer^[1]. In recent years, the widespread adoption of low-dose computed tomography screening has led to a substantial increase in the detection rate of stage IA LUAD. Notably, several recent studies have indicated that the 5-year recurrence-free survival rate for this stage is approximately 90%^[2–4]. Since patients expect surgical treatment to achieve a cure, they find recurrence and metastasis difficult to accept.

Over the past few decades, many researchers have studied the prognosis of patients with stage IA lung cancer. Some clinicopathological variables have been reported to be associated with the prognosis of stage IA LUAD^[5–8]. At the same time, clinical prediction models to predict the prognosis of stage IA LUAD have been established^[9–12]. We found that the prognosis of early-stage lung cancer was discussed mainly using a Cox proportional hazard regression. The Cox model applies to most prognostic studies; however, the application of the Cox model for diseases with low recurrence rates is questionable. Usually, the Cox model assumes that every individual in the study population is susceptible to the event of interest and will eventually experience the event if the follow-up period is long enough. However, the existence of cured patients has often been ignored. For diseases with high recurrence rates, the number of cured patients is small and will not have a significant impact on the experimental results. However, for diseases with low recurrence rates, the existence of cured patients cannot be ignored. For example, the survival curve for most cancers tends to level off after a few years, i.e. a platform effect occurs when the mortality rate of the affected individuals matches the predicted mortality rate of the whole population. These patients are referred to as long-term survivors. For diseases with high recurrence rates, the platform effect appears late, and the use of the conventional Cox model is appropriate. However, the Cox model cannot satisfy the application conditions for stage IA LUAD, which has a low recurrence rate.

In recent years, the mixed cure model has become an important research topic for statisticians^[13–16]. A cure model allows for the direct modelling of the cure rate and the effect of the covariates on the cure rate. In addition, a cure model also allows for the estimation and the inference of the distribution of the survival time for uncured or non-cured patients. In many cases, the same variable affects not only the survival time of patients but also their cure rate. However, in some cases, some covariates have a significant effect on the cure rate of the disease but not on the survival time of the patients. Contrastingly, the other covariates have a more significant effect on the survival time of the patients but little effect on the cure rate of the patients. The Cox model cannot distinguish the influence of these covariates. Therefore, a mixed model for cure was developed to explore the influence of different factors on the survival and cure rates over the disease course.

In this study, we aimed to use a mixed cure model to determine the prognostic factors for stage IA LUAD and perform a sequence analysis on 5%, 10% and 20% presentation subtypes to construct the optimal model and a nomogram to predict the cure rate of patients.

2.1. Patients and data collection

The data of patients with stage IA LUAD who underwent thoracoscopic surgery at Fujian Union Hospital between January 2013 and December 2018 were analyzed. Informed consent was not required as the patient data remained anonymous. All the aspects of this study conformed to the tenets of the Declaration of Helsinki.

The exclusion criteria were: 1) postoperative pathological type of non-adenocarcinoma or mixed carcinoma; 2) postoperative pathology showing adenocarcinoma in situ or minimally invasive adenocarcinoma; 3) patients with other malignant tumors in the past; 4) patients who had received chemotherapy, radiotherapy, targeted therapy or immunotherapy in the past; 5) patients with preoperative blood disease; and 6) patients who died or experienced recurrence within 3 months after surgery due to complications.

The following clinicopathological information was collected: age; sex; smoking history; concomitant disease; preoperative clinical symptoms; surgical methods; tumor location, histological subtype, and diameter; lymphovascular invasion (LVI); preoperative CEA, CA125, and CYFRA21-1 levels; preoperative neutrophil/lymphocyte ratio (NLR) and platelet/lymphocyte ratio (PLR); recurrence time; and time of the last follow-up. Concomitant diseases included hypertension, diabetes, coronary artery disease, and chronic obstructive pulmonary disease. All the surgically resected specimens included in the study were classified by our hospital pathologist according to the 2011 International Lung Cancer Research Association/American Thoracic Society/European Respiratory Society pathological classification. The histological subtypes were classified as lepidic, acinar, papillary, solid, micropapillary, and mucinous adenocarcinoma. Colloid and fetal LUAD were not included in the analysis due to a lack of observable cases. The subtypes are divided into two groups according to the risk of recurrence and metastasis: a low-grade group (lepidic, acinar, and papillary) and a high-grade group (solid, micropapillary, and mucinous adenocarcinoma). Since most LUAD is composed of multiple histological subtypes, according to the 2015 World Health Organization classification of LUAD, the histological subtype that accounts for the largest proportion in postoperative pathological specimens is referred to as the predominant subtype and the secondary component as the presentational subtype. The pathologically predominant subtype is the growth mode with the largest proportion of tumors, and one-by-one analysis was conducted based on the proportion of secondary components in tumor samples (5%, 10%, 20%). LVI was defined as the detection of tumor cells in lymphatic vessels or in the vessel lumen. The surgical methods included a lobectomy and a sub-lobectomy (wedge resection and segmentectomy). The 8th edition of the lung cancer staging system was used for TNM staging.

2.2. Relevant definitions

Recurrence was defined as the appearance of new lesions in the lobe ipsilateral to the primary lesion or in the ipsilateral regional lymph nodes, as demonstrated by relevant histology, cytology or imaging, and relevant clinical evidence showing it as the source of the primary lesion. Metastasis was defined as a new lesion in the lung or an organ outside the lung, as demonstrated by relevant histology, cytology or imaging, and relevant clinical evidence showing it as the source of the primary lesion. Recurrence-free survival (RFS) was defined as the time from the day of discharge to tumor recurrence and metastasis or the last follow-up without recurrence.

2.3. Follow-up

In our department, all postoperative patients were followed up every 3 months for 2 years and then every 6 months. Follow-up information was collected by telephone, and an outpatient evaluation was performed every 6 months after surgery. The follow-up period started on the day of discharge. The end of the follow-up was the last follow-up or the end of follow-up (loss to follow-up, recurrence, or metastasis). The follow-up period for this study ended on May 1, 2022.

2.4. Data analysis and statistical methods

The software packages SPSS.25 and R version 4.3.0 were utilized for statistical analyses. The normal continuous variables were expressed as means ± standard deviations, and the skewed continuous variables as medians (interquartile ranges). The missing values were visualized using the “VIM” package. The missing values were filled in using multiple imputations by SPSS.25 and a reliability analysis was used to select the group of data with the highest Cronbach's alpha for analysis.

A Kaplan–Meier analysis was used to construct the overall survival curve and to test for the presence of long-term survivors and an adequate follow-up time. The Cox proportional hazards cure model is a survival model that includes a cure rate under the assumption that the population contains both uncured and cured patients. It consists of a logistic regression for the cure rate and a Cox regression to estimate the hazard for non-cured patients. Using a mixed cure model that concurrently forecasts the cure rate and risks for uncured patients, a single predictive model for both the cure and risk can be built. We used a Cox proportional hazards cure model, which was defined as:

S(t|x, z) = 1-π(z)+π(z)*S_u(t|x)

where the survival rate partπ(z) represented the probability of no cure, 1-π(z) represented the probability of cure, which was connected by the logistic function, and the incubation period S_u(tΙx) represented the survival function of the uncured patients, which was fitted by the Cox proportional hazards regression. This two-part structure made it easy to consider the influence of the covariance on the cure probability and the survival time distribution of untreated individuals separately, which in turn made the interpretation of the covariant effects attractive and the extension of the mixed cure model to more complex situations easy. The “randomForestSRC” and the “glmnet” packages were used to perform the random forest (RF) algorithm and the least absolute shrinkage and selection operator (lasso), respectively, to filter the variables. The RF algorithm can sample samples and variables to generate a large number of decision trees and predict them in turn. The mode or average of all the decision tree predictions was the final prediction result. Lasso is a data mining technique that effectively reduces the coefficients in linear regression by applying penalty coefficients. This regularization method helped to create a more concise model while mitigating issues of covariance and overfitting. The correlation test between variables was conducted through the R package "correct". The statistical significance was set at P < 0.05.

Using the "rms" and "smcure" packages in R4.1.3, a nomogram of cure probability was developed based on the mixed cure model and used to evaluate the discriminative ability of the mixed cure model using the area under receiver operating characteristic curve (AUC) calculation method of the "evacure" package. For the internal validation of the nomogram's predictive accuracy, bootstrap resamples were performed to determine the bias-corrected C-index. The standard curve was fitted to evaluate the calibration capability of the nomogram model. The closer the calibration curve is to the standard curve, the better the prediction ability of the model.

3.1. Study cohort

Data from 1,064 patients with stage 1A LUAD were analyzed (Table 1). Females (57.8%); patients with no history of smoking (80.5%), no concomitant disease (63.5%), and no preoperative symptoms (81.8%); upper lung location of tumor (60.1%); predominant subtypes (96.8%) in the low-grade group, and presentational subtypes (83.2%) in the low-grade group accounted for the majority of the cohort. Only a few post-operative pathology specimens showed LVI (0.7%); owing to the small number of observable cases, it could not be included in the cure model analysis. There were a few missing values for tumor pathology subtype, and preoperative CEA, CA125, and CYFRA21-1 levels (Supplementary Fig. 1).

Table 1

Participant Characteristics
Variable	Total (n = 1064)
Age (years), median (IQR)	66 (58–72)
CEA (ng/ml), median (IQR)	2.20 (1.40–3.50)
CA125(U/ml), median (IQR)	10.74 (7.58–16.10)
CYFRA21-1(ng/ml), median (IQR)	2.64 (2.00-3.48)
Diameter, median (IQR)	1.5 (1.2-2.0)
NLR, median (IQR)	1.7 (1.3–2.2)
PLR, median (IQR)	121.2 (96.4-153.7)
Sex, n (%)
male	449 (42.2%)
female	615 (57.8%)
Smoking history, n (%)
yes	207 (19.5%)
no	857 (80.5%)
Concomitant disease, n (%)
yes	388 (36.5%)
no	676 (63.5%)
Preoperative symptoms, n (%)
yes	194 (18.2%)
no	870 (81.8%)
Surgical methods, n (%)
lobectomy	154 (14.5%)
sub-lobectomy	910 (85.5%)
LVI, n (%)
yes	7 (0.7%)
no	1057 (99.3%)
Tumor location, n (%)
Upper	639 (60.1%)
Middle	101 (9.5%)
Lower	324 (30.5%)
Predominant subtype, n (%)
Low-grade group	1034 (96.8%)
High-grade group	34 (3.2%)
Presentational subtype (5%), n (%)
Low-grade group	885 (83.2%)
High-grade group	179 (16.8%)
Presentational subtype (10%), n (%)
Low-grade group	939 (88.3%)
High-grade group	125 (11.7%)
Presentational subtype (20%), n (%)
Low-grade group	1022(96.1%)
High-grade group	42 (3.9%)
IQR, interquartile range (25th–75th percentiles); CEA, carcinoembryonic antigen; NLR, preoperative neutrophils/lymphocytes ratio; PLR, preoperative platelet/lymphocyte ratio.

3.2. RF algorithm screening of variables

The RF algorithm was used to filter variables by using the minimum depth algorithm to select the predominant and presentational subtypes; diameter; preoperative CEA, CA125, and CYFRA21-1 levels; LVI; age; NLR; and PLR (Supplementary Tables 1–4, Fig. 1A–B). Given the importance of surgical methods in stage IA LUAD, they were included in the model analysis. The correlation test showed that the collinearity of the 11 variables included in the analysis was not significant (Supplementary Fig. 2).

3.3. Lasso regression screening of variables

As shown in the Fig. 1 (C–D), in order to ensure excellent model performance and an appropriate number of independent variables, the λ value was chosen to be the maximum value within one standard error of the distance minimum mean square error (λ 1se), i.e. in the λ when the value was 0.0259, there were five coefficients in the model that were not zero, which indicated that five independent variables were retained, namely the predominant and presentational subtypes, diameter, LVI, and preoperative CEA level (Supplementary Table 5).

3.4. The cutoff values for the continuity variables

To include time as a variable in the mixed cure model for the calculation of the ideal critical value of the continuous variable, we utilized the X-tile V3.6.1 (a statistical method that employs enumeration)^[17], and separated the patients into two groups based on the critical value. Supplementary Table 6 lists the cutoff values for the continuity variables.

3.5. Survival analysis

Within a median follow-up period of 58 (range, 3–115) months, 73 patients (6.86%) experienced tumor recurrence or metastasis. As shown in Fig. 2, the overall survival curve reached a plateau after 45 months, which may support the hypothesis that the flat part of the RFS curve represented patients who achieve long-term disease control, that is, long-term survivors. We considered that these patients were potentially cured, whereas other patients are considered to have potential recurrence and metastasis. The 3- and 5-year survival rates of the patients were 93.9% and 93.1%, respectively.

3.6. Mixed cure model

As shown in Table 2, the variables screened by the RF algorithm were included in the mixed cure model (using the 5% semi quantitative model records for the presentational subtypes), and the results showed that the predominant subtype (high-grade group, odds ratio [OR]:12.53; 95% confidence interval [CI]:5.51–28.46; P < 0.001), presentational subtype (high-grade group, OR:2.82; 95% CI:1.48–5.39; P = 0.002), and diameter (OR:2.34; 95% CI:1.41–3.87; P < 0.001) had a significant impact on the prognosis of stage IA LUAD in the cure part of the model. In the survival part of the model, the model analysis showed that the only variable that was related to the increased risk of recurrence was the predominant subtype (high-grade group, hazard ratio [HR]:3.69; 95% CI:1.46–9.34; P = 0.01).

Table 2

Results of Cox proportional hazards cure model (RF)
Subgroup	Cure probability model		Failure time distribution model
Subgroup	OR (95%CI)	P-value	HR (95%CI)	P-value
(Intercept)	0.01(0-0.05)	< 0.001
age	1.00(0.97–1.02)	0.76	1.01(0.97–1.05)	0.69
diameter	2.34(1.41–3.87)	< 0.001	0.70(0.34–1.44)	0.33
CEA	1.05(0.99–1.11)	0.11	1.00(0.96–1.04)	0.91
CA125	1.00(0.97–1.02)	0.83	1.00(0.96–1.03)	0.83
CYFRA21-1	1.05(0.89–1.24)	0.56	1.04(0.84–1.29)	0.70
NLR	0.98(0.83–1.16)	0.85	1.12(0.75–1.78)	0.57
PLR	1.00(1.00-1.01)	0.38	1.00(0.99–1.01)	0.74
surgical methods		0.19
lobectomy	reference		reference	0.94
sub-lobectomy	1.69(0.77–3.68)		0.96(0.33–2.79)
predominant subtype		< 0.001		0.01
low grade group	reference		reference
high grade group	12.53(5.51–28.46)		3.69(1.46–9.34)
presentational subtype(5%)		0.002		0.90
low grade group	reference		reference
high grade group	2.82(1.48–5.39)		0.95(0.47–1.94)
OR, odds ratio;HR, Hazard Ratio; CI, confidence interval;

3.7. Estimated survival curves from the Cox proportional hazards cure model

Using the variables selected through the RF algorithm and lasso regression, we applied a mixed cure model to estimate the RFS. A sequential analysis of the presentational subtypes at the 5%, 10%, and 20% levels revealed no significant impact on the postoperative recurrence and metastasis (Tables 2, Supplementary Table 7–11). Among the variables analyzed in the mixed cure model, only the predominant subtype exhibited differences, with a significantly higher HR for the high-grade group compared to the low-grade group. Similarly, the survival curve analysis, excluding long-term survivors, indicated lower survival rates among patients with the predominant subtype in the high-grade group. Due to the limited number of LVI cases, they were not included in the cure model analysis. Nonetheless, the survival curve demonstrated lower survival rates for patients with postoperative pathology indicating LVI (Fig. 3A–M).

3.8 Selection of histologic features in the model of the predominant subtype plus the presentational subtype (with a cutoff of 5%, 10% or 20%)

Supplementary table 12 shows the combined analysis of the variables screened by the RF algorithm and lasso regression, focusing on the presentational subtypes at the 5%, 10%, and 20% levels. We developed six mixed cure models and evaluated their discriminative ability using the AUC, C-index, and K-index. These results showed no significant differences in the discriminative ability among the cure models.

In addition, we used the C-index and the bootstrap-corrected C-index to evaluate the cure part of the mix cure model independently (Supplementary Table 13). Notably, all six models showed an excellent discriminative ability with no significant differences being observed. The calibration plots showed excellent agreement between the nomogram predictions and the actual observations for the cure rate (Supplementary Fig. 3). Therefore, a DCA was performed to determine the clinical usefulness of the nomogram (Fig. 3O). The results showed a good clinical applicability of the nomogram in predicting the patients' cure rate due to a good net benefit with wide and practical ranges of threshold probabilities.

3.9. Nomograms

The discriminant and calibration abilities of the model were good and similar. Considering the practicality of the predictive model, we included the predominant subtype, presentational subtype (10%), diameter, preoperative CEA level, and surgical methods in the nomogram model.

The nomogram comprised five variables and the score axis corresponding to each variable. The score on the nomogram axis corresponded to the state of each variable. The total score equaled the sum of all the variable scores. On the prediction line below the axis, the corresponding predicted cure rate coefficients are shown (Fig. 4).

The C-index for the nomogram was 0.781 (95% CI: 0.730–0.831), and the bootstrap-corrected C-index was 0.768.

Our research showed that most patients with stage IA LUAD were identified through physical examination. Since this research was carried out in developing countries and considering the more advanced medical security in developed countries, this result is credible. In view of this, unlike for most malignant tumors, most of the patients received surgical treatment for the purpose of cure. Therefore, although the prognosis of this disease was good, once the disease relapsed, it was difficult for patients to accept it psychologically. At the same time, during follow-up, it was difficult for us to answer the question, "How likely is my cure rate?" This was the original intention of the present study. The overall survival curve showed that the survival curve reached a stable state 45 months after discharge, and the 5-year RFS rate did not change since then, which indicated the existence of long-term survivors. Although this definition did not necessarily mean that the patient had been cured, it was the definition of a cured population because the mortality of patients was no longer higher than that of the general population, which was the basis for the development of a cure model.

In clinical practice, TNM staging is usually used to judge the prognosis of tumors, although it has considerable limitations in early lung cancer. In the past few decades, researchers have mostly used the Cox regression model to study the factors related to the prognosis of early lung cancer. However, this was not reasonable because it ignored the existence of long-term survivors. Although statisticians realized this problem at the beginning of the 20th century and proposed a cure model^[18, ^19], owing to the complexity of the algorithm of the cure model and lack of a close integration between statistics and clinical medicine, the use of this model in clinical research is still uncommon. Our study primarily involved patients with stage IA LUAD because previous relevant reports have suggested that this disease has a good prognosis and is more suitable for a mixed cure model. In this study, we screened 10 variables using the RF algorithm: the predominant and presentational subtypes; diameter; preoperative CEA, CA125, and CYFRA21-1 levels; LVI; age; NLR; and PLR. In addition, we used lasso regression to screen five variables: the predominant and presentational subtypes, diameter, LVI, and preoperative CEA level. Considering the clinical controversy about whether lobectomy is necessary for the peripheral stage IA LUAD, we included the surgical method as a variable in the analysis. The results showed that the surgical method had no significant effect on both the cure and survival rates. Due to the overall investigation of the prognostic factors in this study, there was no more detailed subgroup analysis of the surgical methods, as it was unnecessary. Recent research results from JCOG1211 and JCOG0804 have also confirmed that there were no significant differences between a lobectomy and sub-lobectomy in terms of the 5-year RFS of patients, and furthermore a lymph node resection was not a risk factor that affected patients' prognosis^[2, ^4].

Predicting the prognosis of patients based solely on tumor diameter is a convenient but inaccurate approach. With the deepening of research, an increasing number of variables have been considered to have a significant impact on the prognosis of patients, among which the tumor histological subtypes have been widely accepted as risk factors for prognosis. LUAD is histologically heterogeneous, showing a combination of multiple subtypes and proportions. The previous approach, which only classified the predominant subtype, had a clear disadvantage because some subtypes, although representing only a small proportion of the tumor, were associated with a poor prognosis^[20–22]. The grading scheme proposed by the IASLC Pathology Committee is based on a combination of the predominant subtype and high-grade group presentational subtype (micropapillary, solid, sieve, and complex glandular) if they represented at least 20% of the tumor^[23]. This is an effective grading scheme; however, considering the good prognosis of stage IA LUAD, which is significantly different from other stages of adenocarcinoma, we explored the 5%, 10%, and 20% incremental recording modes for the high-grade group presentational subtype. Research has shown that the predominant and presentational subtypes are prognostic factors that affect the cure rate of patients. However, from a survival perspective, only the predominant subtype seemed to have an impact on recurrence and metastasis among patients. With regard to the presentational subtype, irrespective of the incremental model used for the analysis, after excluding the existence of long-term survivors, there was no significant impact on the survival rate of patients. It is worth noting that this was not the result for all the stages of LUAD. Therefore, it may have been the case only because the patients with stage IA LUAD have a higher 5-year RFS and comprise a higher proportion of long-term survivors.

Moreover, LVI was too low to be included in the model analysis, and the fitted survival curve suggested that the survival rate of patients with LVI indicated by pathology after surgery was low. In our study, although there were only seven patients with LVI, four patients had relapses and metastases, suggesting that the existence of LVI cannot be ignored. LVI is considered a precursor to tumor metastasis. The presence of tumor cells in blood vessels allows tumors to escape from the primary site. Although LVI has not yet been included in the TNM staging of lung cancer, in previous studies, the existence of LVI was considered generally to be a negative factor for recurrence and metastasis^[24]. Previous studies have reported a significant correlation between LVI and high-grade pathologic subtypes^[25–27]. Due to the low incidence of LVI in the patients from our center and the lack of clear correlation observed, LVI is more likely to be an independent risk factor.

4.1. Limitations

Our study had a few shortcomings. Given that this was a single-center retrospective study, the sample size was not typical of the entire population. Nonetheless, a few parameters were well-controlled. For example, this study only included patients with stage IA LUAD, and all the operations were performed by senior doctors with extensive clinical experience in our department. Although the cure model is not widely used in clinical practice for objective reasons, it has been recognized by the statistical community. We believe that with the development of related algorithms in the future, the cure model can become another effective tool in the field of prognosis analysis.

Through the analysis of the cure model, this study confirmed that the predominant subtype, presentational subtype, and diameter are prognostic factors affecting the cure of patients with stage IA LUAD; however, only the predominant subtype had an impact on the recurrence and metastasis among patients. A sequential analysis of the presentational subtype showed no significant effect on survival (P > 0.05). Regardless of the recording mode, no significant improvement was observed in the discriminative ability of the model. The reason that the findings of this study differs from those of previous studies may be that the cure model considers the existence of long-term survivors. We also built a predictive model to answer the question of a patient’s probability of a cure.

Ethics approval and Consent to participate

The study was approved by the Ethics Committee of Fujian Medical University Union Hospital (No. 2022KY006). The requirement for informed consent was waived by the Ethics Committee of Fujian Medical University Union Hospital due to the retrospective nature of the study.

Consent for publication

Not applicable.

Availability of data and materials

All data relevant to the results are included in this manuscript. Further information on the raw data can be obtained upon request from the corresponding author.

Competing interests

None.

Funding

This work was supported by Key Laboratory of Cardio-Thoracic Surgery, Fujian Medical University, Fujian Province University.

Acknowledgments

We would like to thank Editage (www.editage.cn) for their English language editing services.

Author contributions

(I) Conception and design: Mingqiang Kang and Zhixin Huang; (II) Administrative support: Mingqiang Kang; (III) Provision of study materials or patients: Zhixin Huang; (IV) Collection and assembly of data: Zhixin Huang and Dinghang Chen; (V) Data analysis and interpretation: Zhixin Huang; (VI) Manuscript writing: All authors; (VII) Final approval of manuscript: All authors.

Sung H, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021;71(3):209–49.
Saji H, et al. Segmentectomy versus lobectomy in small-sized peripheral non-small-cell lung cancer (JCOG0802/WJOG4607L): a multicentre, open-label, phase 3, randomised, controlled, non-inferiority trial. Lancet. 2022;399(10335):1607–17.
Suzuki K, et al. A nonrandomized confirmatory phase III study of sublobar surgical resection for peripheral ground glass opacity dominant lung cancer defined with thoracic thin-section computed tomography (JCOG0804/WJOG4507L). J Clin Oncol. 2017;35(15suppl):8561–1.
Aokage K, et al. Segmentectomy for ground-glass-dominant lung cancer with a tumour diameter of 3 cm or less including ground-glass opacity (JCOG1211): a multicentre, single-arm, confirmatory, phase 3 trial. Lancet Respir Med. 2023;11(6):540–9.
Wang T, et al. Radiomics for Survival Risk Stratification of Clinical and Pathologic Stage IA Pure-Solid Non-Small Cell Lung Cancer. Radiology. 2022;302(2):425–34.
Zhao Y, et al. Lobe-specific Lymph Node Dissection in Clinical Stage IA Solid-dominant Non-small-cell Lung Cancer: A Propensity Score Matching Study. Clin Lung Cancer. 2021;22(2):e201–10.
Abughararah TZ, et al. Lobe-specific lymph node dissection in stage IA non-small-cell lung cancer: a retrospective cohort study. Eur J Cardiothorac Surg. 2021;59(4):783–90.
Saw S, et al. Association of Clinicopathologic and Molecular Tumor Features With Recurrence in Resected Early-Stage Epidermal Growth Factor Receptor-Positive Non-Small Cell Lung Cancer. JAMA Netw Open. 2021;4(11):e2131892.
Xu W, et al. A 10-Gene Yin Yang Expression Ratio Signature for Stage IA and IB Non-Small Cell Lung Cancer. J Thorac Oncol. 2016;11(12):2150–60.
Wo Y, et al. Development and External Validation of a Nomogram for Predicting Survival in Patients With Stage IA Non-small Cell Lung Cancer ≤ 2 cm Undergoing Sublobectomy. Front Oncol. 2019;9:1385.
Merritt RE, et al. Nomograms for Predicting Overall and Recurrence-free Survival From Pathologic Stage IA and IB Lung Cancer After Lobectomy. Clin Lung Cancer. 2021;22(4):e574–83.
Huang Z et al. Nomogram for predicting recurrence and metastasis of stage IA lung adenocarcinoma treated by video-assisted thoracoscopic surgery lobectomy. Asian J Surg, 2022.
Cvancarova M, et al. Proportion cured models applied to 23 cancer sites in Norway. Int J Cancer. 2013;132(7):1700–10.
Othus M, et al. Accounting for Cured Patients in Cost-Effectiveness Analysis. Value Health. 2017;20(4):705–9.
Pedrosa-Laza M, López-Cheda A, Cao R. Cure models to estimate time until hospitalization due to COVID-19: A case study in Galicia (NW Spain). Appl Intell (Dordr). 2022;52(1):794–807.
Beesley LJ, Taylor J. EM algorithms for fitting multistate cure models. Biostatistics. 2019;20(3):416–32.
RL C, D. M and, DL R. X-tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization. Clin cancer research: official J Am Association Cancer Res. 2004;10(21):7252–9.
Boag JW. Maximum Likelihood Estimates of the Proportion of Patients Cured by Cancer Therapy. J Royal Stat Soc Ser B: Methodological. 1948;11(1):15–53.
Berkson J, Gage RP. Survival Curve for Cancer Patients Following Treatment. J Am Stat Assoc. 1952;47(259):501–15.
Wang W, et al. Both the presence of a micropapillary component and the micropapillary predominant subtype predict poor prognosis after lung adenocarcinoma resection: a meta-analysis. J Cardiothorac Surg. 2020;15(1):154.
Yuan Y, et al. Presence of micropapillary and solid patterns are associated with nodal upstaging and unfavorable prognosis among patient with cT1N0M0 lung adenocarcinoma: a large-scale analysis. J Cancer Res Clin Oncol. 2018;144(4):743–9.
Zhao Y, et al. Minor Components of Micropapillary and Solid Subtypes in Lung Adenocarcinoma are Predictors of Lymph Node Metastasis and Poor Prognosis. Ann Surg Oncol. 2016;23(6):2099–105.
Moreira AL, et al. A Grading System for Invasive Pulmonary Adenocarcinoma: A Proposal From the International Association for the Study of Lung Cancer Pathology Committee. J Thorac Oncol. 2020;15(10):1599–610.
Ruffini E, et al. Significance of the presence of microscopic vascular invasion after complete resection of Stage I-II pT1-T2N0 non-small cell lung cancer and its relation with T-Size categories: did the 2009 7th edition of the TNM staging system miss something? J Thorac Oncol. 2011;6(2):319–26.
Fujikawa R, et al. Clinicopathologic and Genotypic Features of Lung Adenocarcinoma Characterized by the International Association for the Study of Lung Cancer Grading System. J Thorac Oncol. 2022;17(5):700–7.
Ahn B, et al. Clinicopathologic and genomic features of high-grade pattern and their subclasses in lung adenocarcinoma. Lung Cancer. 2022;170:176–84.
Saruwatari K, et al. Aggressive tumor microenvironment of solid predominant lung adenocarcinoma subtype harboring with epidermal growth factor receptor mutations. Lung Cancer. 2016;91:7–14.

No competing interests reported.

Download PDF

Journal Publication

published 31 Oct, 2023

Read the published version in BMC Pulmonary Medicine →

Editorial decision: Major revision
17 Sep, 2023
Reviews received at journal
12 Sep, 2023
Reviewers agreed at journal
31 Aug, 2023
Reviewers invited by journal
12 Aug, 2023
Editor assigned by journal
12 Aug, 2023
Editor invited by journal
27 Jul, 2023
Submission checks completed at journal
27 Jul, 2023
First submitted to journal
06 Jul, 2023

You are reading this latest preprint version

Estimating the Cure Proportion in Stage IA Lung Adenocarcinoma: A population-based study

Status:

Journal Publication

Version 1

Abstract

Objectives

Materials and Methods

Results

Conclusion

Figures

1. Introduction

2. Material and methods

2.1. Patients and data collection

2.2. Relevant definitions

2.3. Follow-up

2.4. Data analysis and statistical methods

3. Results

3.1. Study cohort

3.2. RF algorithm screening of variables

3.3. Lasso regression screening of variables

3.4. The cutoff values for the continuity variables

3.5. Survival analysis

3.6. Mixed cure model

3.7. Estimated survival curves from the Cox proportional hazards cure model

3.9. Nomograms

4. Discussion

4.1. Limitations

5. Conclusions

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1