Risk prediction models for selection of lung cancer screening candidates: A retrospective validation study

The Impact of Model Assumptions on Personalized Lung Cancer Screening Recommendations

BACKGROUND

Recommendations regarding personalized lung cancer screening are being informed by natural-history modeling. Therefore, understanding how differences in model assumptions affect model-based personalized screening recommendations is essential.

DESIGN

Five Cancer Intervention and Surveillance Modeling Network (CISNET) models were evaluated. Lung cancer incidence, mortality, and stage distributions were compared across 4 theoretical scenarios to assess model assumptions regarding 1) sojourn times, 2) stage-specific sensitivities, and 3) screening-induced lung cancer mortality reductions. Analyses were stratified by sex and smoking behavior.

RESULTS

Most cancers had sojourn times <5 y (model range [MR]; lowest to highest value across models: 83.5%-98.7% of cancers). However, cancer aggressiveness still varied across models, as demonstrated by differences in proportions of cancers with sojourn times <2 y (MR: 42.5%-64.6%) and 2 to 4 y (MR: 28.8%-43.6%). Stage-specific sensitivity varied, particularly for stage I (MR: 31.3%-91.5%). Screening reduced stage IV incidence in most models for 1 y postscreening; increased sensitivity prolonged this period to 2 to 5 y. Screening-induced lung cancer mortality reductions among lung cancers detected at screening ranged widely (MR: 14.6%-48.9%), demonstrating variations in modeled treatment effectiveness of screen-detected cases. All models assumed longer sojourn times and greater screening-induced lung cancer mortality reductions for women. Models assuming differences in cancer epidemiology by smoking behaviors assumed shorter sojourn times and lower screening-induced lung cancer mortality reductions for heavy smokers.

CONCLUSIONS

Model-based personalized screening recommendations are primarily driven by assumptions regarding sojourn times (favoring longer intervals for groups more likely to develop less aggressive cancers), sensitivity (higher sensitivities favoring longer intervals), and screening-induced mortality reductions (greater reductions favoring shorter intervals).

IMPLICATIONS

Models suggest longer screening intervals may be feasible and benefits may be greater for women and light smokers.

HIGHLIGHTS

Natural-history models are increasingly used to inform lung cancer screening, but causes for variations between models are difficult to assess.This is the first evaluation of these causes and their impact on personalized screening recommendations through easily interpretable metrics.Models vary regarding sojourn times, stage-specific sensitivities, and screening-induced lung cancer mortality reductions.Model outcomes were similar in predicting greater screening benefits for women and potentially light smokers. Longer screening intervals may be feasible for women and light smokers.

A dynamic online nomogram for predicting the heterogeneity trajectories of frailty among elderly gastric cancer survivors

BACKGROUND

Frailty is very common among older people with gastric cancer and seriously affects their prognosis. The development of frailty is continuous and dynamic, increasing the difficulty and burden of care.

OBJECTIVES

The aims of this study were to delineate the developmental trajectory of frailty in older people with gastric cancer 1 year after surgery, identify heterogeneous frailty trajectories, and further explore their predictors to construct a nomogram for prediction.

DESIGN

We conducted a prospective longitudinal observation study. Clinical evaluation and data collection were performed at discharge, and at 1, 3, 6, and 12 months.

SETTING AND PARTICIPANTS

This study was conducted in a tertiary hospital and 381 gastric cancer patients over 60 years who underwent radical gastrectomy completed the 1-year follow-up.

METHODS

A growth mixture model (GMM) was used to delineate the frailty trajectories, and identify heterogeneous trajectories. A regression model was performed to determine their predictors and further construct a nomogram based on the predictors. Bootstrap with 1000 resamples was used for internal validation of nomogram, a receiver operating characteristic (ROC) curve to evaluate discrimination, calibration curves to evaluate calibration and decision curve analysis (DCA) to evaluate the clinical value.

RESULTS

GMM identified three classes of frailty trajectories: "frailty improving", "frailty persisting" and "frailty deteriorating". The latter two were referred to as heterogeneous frailty trajectories. Regression analysis showed 8 independent predictors of heterogeneous frailty trajectories and a nomogram was constructed based on these predictors. The area under ROC curve (AUC) of the nomogram was 0.731 (95 % CI = 0.679-0.781), the calibration curves demonstrated that probabilities predicted by the nomogram agreed well with the actual observation with a mean absolute error of 0.025, and the DCA of nomogram indicated that the net benefits were higher than that of the other eight single factors.

CONCLUSIONS

Older gastric cancer patients have heterogeneous frailty trajectories of poor prognosis during one-year postoperative survival. Therefore, early assessment to predict the occurrence of heterogeneous frailty trajectories is essential to improve the outcomes of elderly gastric cancer patients. Scientific and effective frailty interventions should be further explored in the future to improve the prognosis of older gastric cancer patients.

CONTRIBUTION OF THE PAPER STATEMENTS

This study constructed a static and dynamic online nomogram with good discrimination and calibration, which can help to screen high-risk patients, implement preoperative risk stratification easily and promote the rational allocation of medical resources greatly.

REGISTRATION

ClinicalTrials.gov (Number: NCT05982899).

TWEETABLE ABSTRACT

Our findings identified three frailty trajectories and constructed a nomogram to implement preoperative risk stratification and improve patient outcomes.

Safety net hospital risk model demonstrates stronger, population-specific applicability in characterizing lung cancer risk

Background

Determining lung cancer (LC) risk using personalized risk stratification may improve screening effectiveness. While the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO) is a well-established stratification model for LC screening, it was derived from a predominantly Caucasian population and its effectiveness in a safety net hospital (SNH) population is unknown. We have developed a model more tailored to the SNH population and compared its performance to the PLCO model in a SNH setting.

Methods

Retrospective dataset was compiled from patients screened for LC at SNH from 2015 to 2019. Descriptive statistics were calculated using the following variables: age, sex, race, education, body mass index (BMI), smoking history, personal cancer history, family LC history, chronic obstructive pulmonary disease (COPD), and emphysema. Variables distribution was compared using t- and chi-square tests. LC risk scores were calculated using SNH and PLCO models and categorized as low (scores <0.65%), moderate (0.65-1.49%), and high (>1.5%). Linear regression was applied to evaluate the relationship between models and covariates.

Results

Of 896 individuals, 38 were diagnosed with LC. Data reflected the SNH patient demographics, which predominantly were African American (53.5%), current smokers (69.9%), and with emphysema (70.1%). Among the non-LC cohort, SNH model most frequently categorized patients as low risk, while PLCO model most frequently classified patients as moderate risk. Among the LC cohort, there was no significant difference between mean scores or risk stratification. SNH model showed 92.1% sensitivity and 96.8% specificity while PLCO model showed 89.4% sensitivity and 26.1% specificity. Emphysema demonstrated a strong association in SNH model (P<0.001) while race showed no relation.

Conclusions

SNH model demonstrated greater specificity for characterizing LC risk in a SNH population. The results demonstrated the importance of study sample representation when identifying risk factors in a stratification model.

Personalised follow-up and management schema for patients with screen-detected pulmonary nodules: A dynamic modelling study

BACKGROUND

Selecting the time target for follow-up testing in lung cancer screening is challenging. We aim to devise dynamic, personalized lung cancer screening schema for patients with pulmonary nodules detected through low-dose computed tomography.

METHODS

We developed and validated dynamic models using data of pulmonary nodule patients (aged 55-74 years) from the National Lung Screening Trial. We predicted patient-specific risk profiles at baseline (R0) and updated the risk evaluation results in repeated screening rounds (R1 and R2). We used risk cutoffs to optimize time-dependent sensitivity at an early decision point (3 months) and time-dependent specificity at a late decision point (1 year).

RESULTS

In validation, area under receiver operating characteristic curve for predicting 12-month lung cancer onset was 0.867 (95 % confidence interval: 0.827-0.894) and 0.807 (0.765-0.948) at R0 and R1-R2, respectively. The personalized schema, compared with National Comprehensive Cancer Network (NCCN) guideline and Lung-RADS, yielded lower rates of delayed diagnosis (1.7% vs. 1.7% vs. 6.9 %) and over-testing (4.9% vs. 5.6% vs. 5.6 %) at R0, and lower rates of delayed diagnosis (0.0% vs. 18.2% vs. 18.2 %) and over-testing (2.6% vs. 8.3% vs. 7.3 %) at R2. Earlier test recommendation among cancer patients was more frequent using the personalized schema (vs. NCCN: 29.8% vs. 20.9 %, p = 0.0065; vs. Lung-RADS: 33.2% vs. 22.8 %, p = 0.0025), especially for women, patients aged ≥65 years, and part-solid or non-solid nodules.

CONCLUSIONS

The personalized schema is easy-to-implement and more accurate compared with rule-based protocols. The results highlight value of personalized approaches in realizing efficient nodule management.

Clinical Utility of a CT-based AI Prognostic Model for Segmentectomy in Non-Small Cell Lung Cancer

Background Currently, no tool exists for risk stratification in patients undergoing segmentectomy for non-small cell lung cancer (NSCLC). Purpose To develop and validate a deep learning (DL) prognostic model using preoperative CT scans and clinical and radiologic information for risk stratification in patients with clinical stage IA NSCLC undergoing segmentectomy. Materials and Methods In this single-center retrospective study, transfer learning of a pretrained model was performed for survival prediction in patients with clinical stage IA NSCLC who underwent lobectomy from January 2008 to March 2017. The internal set was divided into training, validation, and testing sets based on the assignments from the pretraining set. The model was tested on an independent test set of patients with clinical stage IA NSCLC who underwent segmentectomy from January 2010 to December 2017. Its prognostic performance was analyzed using the time-dependent area under the receiver operating characteristic curve (AUC), sensitivity, and specificity for freedom from recurrence (FFR) at 2 and 4 years and lung cancer-specific survival and overall survival at 4 and 6 years. The model sensitivity and specificity were compared with those of the Japan Clinical Oncology Group (JCOG) eligibility criteria for sublobar resection. Results The pretraining set included 1756 patients. Transfer learning was performed in an internal set of 730 patients (median age, 63 years [IQR, 56-70 years]; 366 male), and the segmentectomy test set included 222 patients (median age, 65 years [IQR, 58-71 years]; 114 male). The model performance for 2-year FFR was as follows: AUC, 0.86 (95% CI: 0.76, 0.96); sensitivity, 87.4% (7.17 of 8.21 patients; 95% CI: 59.4, 100); and specificity, 66.7% (136 of 204 patients; 95% CI: 60.2, 72.8). The model showed higher sensitivity for FFR than the JCOG criteria (87.4% vs 37.6% [3.08 of 8.21 patients], P = .02), with similar specificity. Conclusion The CT-based DL model identified patients at high risk among those with clinical stage IA NSCLC who underwent segmentectomy, outperforming the JCOG criteria. © RSNA, 2024 Supplemental material is available for this article.

Multiomics-Based Feature Extraction and Selection for the Prediction of Lung Cancer Survival

Lung cancer is a global health challenge, hindered by delayed diagnosis and the disease's complex molecular landscape. Accurate patient survival prediction is critical, motivating the exploration of various -omics datasets using machine learning methods. Leveraging multi-omics data, this study seeks to enhance the accuracy of survival prediction by proposing new feature extraction techniques combined with unbiased feature selection. Two lung adenocarcinoma multi-omics datasets, originating from the TCGA and CPTAC-3 projects, were employed for this purpose, emphasizing gene expression, methylation, and mutations as the most relevant data sources that provide features for the survival prediction models. Additionally, gene set aggregation was shown to be the most effective feature extraction method for mutation and copy number variation data. Using the TCGA dataset, we identified 32 molecular features that allowed the construction of a 2-year survival prediction model with an AUC of 0.839. The selected features were additionally tested on an independent CPTAC-3 dataset, achieving an AUC of 0.815 in nested cross-validation, which confirmed the robustness of the identified features.

Lung Cancer Risk Prediction Models for Asian Ever-Smokers

INTRODUCTION

Although lung cancer prediction models are widely used to support risk-based screening, their performance outside Western populations remains uncertain. This study aims to evaluate the performance of 11 existing risk prediction models in multiple Asian populations and to refit prediction models for Asians.

METHODS

In a pooled analysis of 186,458 Asian ever-smokers from 19 prospective cohorts, we assessed calibration (expected-to-observed ratio) and discrimination (area under the receiver operating characteristic curve [AUC]) for each model. In addition, we developed the "Shanghai models" to better refine risk models for Asians on the basis of two well-characterized population-based prospective cohorts and externally validated them in other Asian cohorts.

RESULTS

Among the 11 models, the Lung Cancer Death Risk Assessment Tool yielded the highest AUC (AUC [95% confidence interval (CI)] = 0.71 [0.67-0.74] for lung cancer death and 0.69 [0.67-0.72] for lung cancer incidence) and the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial Model had good calibration overall (expected-to-observed ratio [95% CI] = 1.06 [0.90-1.25]). Nevertheless, these models substantially underestimated lung cancer risk among Asians who reported less than 10 smoking pack-years or stopped smoking more than or equal to 20 years ago. The Shanghai models were found to have marginal improvement overall in discrimination (AUC [95% CI] = 0.72 [0.69-0.74] for lung cancer death and 0.70 [0.67-0.72] for lung cancer incidence) but consistently outperformed the selected Western models among low-intensity smokers and long-term quitters.

CONCLUSIONS

The Shanghai models had comparable performance overall to the best existing models, but they improved much in predicting the lung cancer risk of low-intensity smokers and long-term quitters in Asia.

Cost-effectiveness of risk-based low-dose computed tomography screening for lung cancer in Switzerland

Throughout Europe, computed tomography (CT) screening for lung cancer is in a phase of clinical implementation or reimbursement evaluation. To efficiently select individuals for screening, the use of lung cancer risk models has been suggested, but their incremental (cost-)effectiveness relative to eligibility based on pack-year criteria has not been thoroughly evaluated for a European setting. We evaluate the cost-effectiveness of pack-year and risk-based screening (PLCOm2012 model-based) strategies for Switzerland, which aided in informing the recommendations of the Swiss Cancer Screening Committee (CSC). We use the MISCAN (MIcrosimulation SCreening ANalysis)-Lung model to estimate benefits and harms of screening among individuals born 1940 to 1979 in Switzerland. We evaluate 1512 strategies, differing in the age ranges employed for screening, the screening interval and the strictness of the smoking requirements. We estimate risk-based strategies to be more cost-effective than pack-year-based screening strategies. The most efficient strategy compliant with CSC recommendations is biennial screening for ever-smokers aged 55 to 80 with a 1.6% PLCOm2012 risk. Relative to no screening this strategy is estimated to reduce lung cancer mortality by 11.0%, with estimated costs per Quality-Adjusted Life-Year (QALY) gained of €19 341, and a €1.990 billion 15-year budget impact. Biennial screening ages 55 to 80 for those with 20 pack-years shows a lower mortality reduction (10.5%) and higher cost per QALY gained (€20 869). Despite model uncertainties, our estimates suggest there may be cost-effective screening policies for Switzerland. Risk-based biennial screening ages 55 to 80 for those with ≥1.6% PLCOm2012 risk conforms to CSC recommendations and is estimated to be more efficient than pack-year-based alternatives.

Lung nodule malignancy classification with associated pulmonary fibrosis using 3D attention-gated convolutional network with CT scans

BACKGROUND

Chest Computed tomography (CT) scans detect lung nodules and assess pulmonary fibrosis. While pulmonary fibrosis indicates increased lung cancer risk, current clinical practice characterizes nodule risk of malignancy based on nodule size and smoking history; little consideration is given to the fibrotic microenvironment.

PURPOSE

To evaluate the effect of incorporating fibrotic microenvironment into classifying malignancy of lung nodules in chest CT images using deep learning techniques.

MATERIALS AND METHODS

We developed a visualizable 3D classification model trained with in-house CT dataset for the nodule malignancy classification task. Three slightly-modified datasets were created: (1) nodule alone (microenvironment removed); (2) nodule with surrounding lung microenvironment; and (3) nodule in microenvironment with semantic fibrosis metadata. For each of the models, tenfold cross-validation was performed. Results were evaluated using quantitative measures, such as accuracy, sensitivity, specificity, and area-under-curve (AUC), as well as qualitative assessments, such as attention maps and class activation maps (CAM).

RESULTS

The classification model trained with nodule alone achieved 75.61% accuracy, 50.00% sensitivity, 88.46% specificity, and 0.78 AUC; the model trained with nodule and microenvironment achieved 79.03% accuracy, 65.46% sensitivity, 85.86% specificity, and 0.84 AUC. The model trained with additional semantic fibrosis metadata achieved 80.84% accuracy, 74.67% sensitivity, 84.95% specificity, and 0.89 AUC. Our visual evaluation of attention maps and CAM suggested that both the nodules and the microenvironment contributed to the task.

CONCLUSION

The nodule malignancy classification performance was found to be improving with microenvironment data. Further improvement was found when incorporating semantic fibrosis information.

Serum-based biomarkers associated with lung cancer risk and cause-specific mortality in the German randomized Lung Cancer Screening Intervention (LUSI) trial

Background

Lung cancer (LC) screening can be optimized using individuals' estimated risks of having a detectable lung tumor, as well as of mortality risk by competing causes, to guide decisions on screening eligibility, ideal screening intervals and stopping ages. Besides age, sex and smoking history, blood-based biomarkers may be used to improve the assessment of LC risk and risk of mortality by competing causes.

Methods

In the German randomized Lung Screening Intervention Trial (LUSI), we measured growth/differentiation factor-15 (GDF-15), interleukin-6 (IL-6), C-reactive protein (CRP) and N-terminal pro-brain natriuretic protein (NT-proBNP), in blood serum samples collected at start of the trial. Participants in the computed tomography (CT)-screening arm also had a pulmonary function test. Regression models were used to examine these markers as predictors for impaired lung function, LC risk and mortality due to LC or other causes, independently of age, sex and smoking history.

Results

Our models showed increases in LC risk among participants with elevated serum levels of GDF-15 [odds ratio (OR)Q4-Q1 =2.47, 95% confidence interval (CI): 1.49-4.26], IL-6 [ORQ4-Q1 =2.36 (1.43-4.00)] and CRP [ORQ4-Q1 =1.81 (1.08-2.75)]. Likewise, proportional hazards models showed increased risks for LC-related mortality, hazard ratio (HR)Q4-Q1 of 4.63 (95% CI: 2.13-10.07) for GDF-15, 3.56 (1.72-7.37) for IL-6 and 2.34 (1.24-4.39) for CRP. All four markers were associated with increased risk of mortality by causes other than LC, with strongest associations for GDF-15 [HRQ4-Q1 =3.04 (2.09-4.43)] and IL-6 [HRQ4-Q1 =2.98 (2.08-4.28)]. Significant associations were also observed between IL-6, CRP, GDF-15 and impaired pulmonary function [chronic obstructive pulmonary disease (COPD), preserved ratio impaired spirometry (PRISm)]. Multi-marker models identified GDF-15 and IL-6 as joint risk predictors for risk of LC diagnosis, without further discrimination by CRP or NT-proBNP. A model based on age, sex, smoking-related variables, GFD-15 and IL-6 provided moderately strong discrimination for prediction of LC diagnoses within 9 years after blood sampling [area under the curve (AUC) =74.3% (57.3-90.2%)], compared to 67.0% (49.3-84.8%) for a model without biomarkers. For mortality by competing causes, a model including biomarkers resulted in an AUC of 76.2% (66.6-85.3%)], compared to 70.0% (60.9-77.9%) a model including age, sex and smoking variables.

Conclusions

Serum GDF-15 and IL-6 may be useful indicators for estimating risks for LC and competing mortality among long-term smokers participating in LC screening, to optimize LC screening strategies.

Risk Model-Based Lung Cancer Screening and Racial and Ethnic Disparities in the US

Importance

The revised 2021 US Preventive Services Task Force (USPSTF) guidelines for lung cancer screening have been shown to reduce disparities in screening eligibility and performance between African American and White individuals vs the 2013 guidelines. However, potential disparities across other racial and ethnic groups in the US remain unknown. Risk model-based screening may reduce racial and ethnic disparities and improve screening performance, but neither validation of key risk prediction models nor their screening performance has been examined by race and ethnicity.

Objective

To validate and recalibrate the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial 2012 (PLCOm2012) model-a well-established risk prediction model based on a predominantly White population-across races and ethnicities in the US and evaluate racial and ethnic disparities and screening performance through risk-based screening using PLCOm2012 vs the USPSTF 2021 criteria.

Design, Setting, and Participants

In a population-based cohort design, the Multiethnic Cohort Study enrolled participants in 1993-1996, followed up through December 31, 2018. Data analysis was conducted from April 1, 2022, to May 19. 2023. A total of 105 261 adults with a smoking history were included.

Exposures

The 6-year lung cancer risk was calculated through recalibrated PLCOm2012 (ie, PLCOm2012-Update) and screening eligibility based on a 6-year risk threshold greater than or equal to 1.3%, yielding similar eligibility as the USPSTF 2021 guidelines.

Outcomes

Predictive accuracy, screening eligibility-incidence (E-I) ratio (ie, ratio of the number of eligible to incident cases), and screening performance (sensitivity, specificity, and number needed to screen to detect 1 lung cancer).

Results

Of 105 261 participants (60 011 [57.0%] men; mean [SD] age, 59.8 [8.7] years), consisting of 19 258 (18.3%) African American, 27 227 (25.9%) Japanese American, 21 383 (20.3%) Latino, 8368 (7.9%) Native Hawaiian/Other Pacific Islander, and 29 025 (27.6%) White individuals, 1464 (1.4%) developed lung cancer within 6 years from enrollment. The PLCOm2012-Update showed good predictive accuracy across races and ethnicities (area under the curve, 0.72-0.82). The USPSTF 2021 criteria yielded a large disparity among African American individuals, whose E-I ratio was 53% lower vs White individuals (E-I ratio: 9.5 vs 20.3; P < .001). Under the risk-based screening (PLCOm2012-Update 6-year risk ≥1.3%), the disparity between African American and White individuals was substantially reduced (E-I ratio: 15.9 vs 18.4; P < .001), with minimal disparities observed in persons of other minoritized groups, including Japanese American, Latino, and Native Hawaiian/Other Pacific Islander. Risk-based screening yielded superior overall and race and ethnicity-specific performance to the USPSTF 2021 criteria, with higher overall sensitivity (67.2% vs 57.7%) and lower number needed to screen (26 vs 30) at similar specificity (76.6%).

Conclusions

The findings of this cohort study suggest that risk-based lung cancer screening can reduce racial and ethnic disparities and improve screening performance across races and ethnicities vs the USPSTF 2021 criteria.

Personalised lung cancer risk stratification and lung cancer screening: do general practice electronic medical records have a role?

BACKGROUND

In the United Kingdom (UK), cancer screening invitations are based on general practice (GP) registrations. We hypothesize that GP electronic medical records (EMR) can be utilised to calculate a lung cancer risk score with good accuracy/clinical utility.

METHODS

The development cohort was Secure Anonymised Information Linkage-SAIL (2.3 million GP EMR) and the validation cohort was UK Biobank-UKB (N = 211,597 with GP-EMR availability). Fast backward method was applied for variable selection and area under the curve (AUC) evaluated discrimination.

RESULTS

Age 55-75 were included (SAIL: N = 574,196; UKB: N = 137,918). Six-year lung cancer incidence was 1.1% (6430) in SAIL and 0.48% (656) in UKB. The final model included 17/56 variables in SAIL for the EMR-derived score: age, sex, socioeconomic status, smoking status, family history, body mass index (BMI), BMI:smoking interaction, alcohol misuse, chronic obstructive pulmonary disease, coronary heart disease, dementia, hypertension, painful condition, stroke, peripheral vascular disease and history of previous cancer and previous pneumonia. The GP-EMR-derived score had AUC of 80.4% in SAIL and 74.4% in UKB and outperformed ever-smoked criteria (currently the first step in UK lung cancer screening pilots).

DISCUSSION

A GP-EMR-derived score may have a role in UK lung cancer screening by accurately targeting high-risk individuals without requiring patient contact.

Integrated plasma and exosome long noncoding RNA profiling is promising for diagnosing non-small cell lung cancer

OBJECTIVES

Non-small cell lung cancer (NSCLC) accounts for more than 80 % of all lung cancers, and its 5-year survival rate can be greatly improved by early diagnosis. However, early diagnosis remains elusive because of the lack of effective biomarkers. In this study, we aimed to develop an effective diagnostic model for NSCLC based on a combination of circulating biomarkers.

METHODS

Tissue-deregulated long noncoding RNAs (lncRNAs) in NSCLC were identified in datasets retrieved from the Gene Expression Omnibus (GEO, n=727) and The Cancer Genome Atlas (TCGA, n=1,135) databases, and their differential expression was verified in paired local plasma and exosome samples from NSCLC patients. Subsequently, LASSO regression was used to screen for biomarkers in a large clinical population, and a logistic regression model was used to establish a multi-marker diagnostic model. The area under the receiver operating characteristic (ROC) curve (AUC), calibration plots, decision curve analysis (DCA), clinical impact curves, and integrated discrimination improvement (IDI) were used to evaluate the efficiency of the diagnostic model.

RESULTS

Three lncRNAs-PGM5-AS1, SFTA1P, and CTA-384D8.35 were consistently expressed in online tissue datasets, plasma, and exosomes from local patients. LASSO regression identified nine variables (Plasma CTA-384D8.35, Plasma PGM5-AS1, Exosome CTA-384D8.35, Exosome PGM5-AS1, Exosome SFTA1P, Log10CEA, Log10CA125, SCC, and NSE) in clinical samples that were eventually included in the multi-marker diagnostic model. Logistic regression analysis revealed that Plasma CTA-384D8.35, exosome SFTA1P, Log10CEA, Exosome CTA-384D8.35, SCC, and NSE were independent risk factors for NSCLC (p<0.01), and their results were visualized using a nomogram to obtain personalized prediction outcomes. The constructed diagnostic model demonstrated good NSCLC prediction ability in both the training and validation sets (AUC=0.97).

CONCLUSIONS

In summary, the constructed circulating lncRNA-based diagnostic model has good NSCLC prediction ability in clinical samples and provides a potential diagnostic tool for NSCLC.

Clinical Adjuncts to Lung Cancer Screening: A Narrative Review

The updated US Preventive Services Task Force guidelines on lung cancer screening have significantly expanded the population of screening eligible adults, among whom the balance of benefits and harms associated with lung cancer screening vary considerably. Clinical adjuncts are additional information and tools that can guide decision-making to optimally screen individuals who are most likely to benefit. Proposed adjuncts include integration of clinical history, risk prediction models, shared-decision-making tools, and biomarker tests at key steps in the screening process. Although evidence regarding their clinical utility and implementation is still evolving, they carry significant promise in optimizing screening effectiveness and efficiency for lung cancer.

CT Imaging With Machine Learning for Predicting Progression to COPD in Individuals at Risk

BACKGROUND

Identifying individuals at risk of progressing to COPD may allow for initiation of treatment to potentially slow the progression of the disease or the selection of subgroups for discovery of novel interventions.

RESEARCH QUESTION

Does the addition of CT imaging features, texture-based radiomic features, and established quantitative CT scan to conventional risk factors improve the performance for predicting progression to COPD in individuals who smoke with machine learning?

STUDY DESIGN AND METHODS

Participants at risk (individuals who currently or formerly smoked, without COPD) from the Canadian Cohort Obstructive Lung Disease (CanCOLD) population-based study underwent CT imaging at baseline and spirometry at baseline and follow-up. Various combinations of CT scan features, texture-based CT scan radiomics (n = 95), and established quantitative CT scan (n = 8), as well as demographic (n = 5) and spirometry (n = 3) measurements, with machine learning algorithms were evaluated to predict progression to COPD. Performance metrics included the area under the receiver operating characteristic curve (AUC) to evaluate the models. DeLong test was used to compare the performance of the models.

RESULTS

Among the 294 at-risk participants who were evaluated (mean age, 65.6 ± 9.2 years; 42% female; mean pack-years, 17.9 ± 18.7), 52 participants (23.7%) in the training data set and 17 participants (23.0%) in the testing data set progressed to spirometric COPD at follow-up (2.5 ± 0.9 years from baseline). Compared with machine learning models with demographics alone (AUC, 0.649), the addition of CT imaging features to demographics (AUC, 0.730; P < .05) or CT imaging features and spirometry to demographics (AUC, 0.877; P < .05) significantly improved the performance for predicting progression to COPD.

INTERPRETATION

Heterogeneous structural changes occur in the lungs of individuals at risk that can be quantified using CT imaging features, and evaluation of these features together with conventional risk factors improves performance for predicting progression to COPD.

Update on Lung Cancer Screening Guideline

Lung cancer screening has been shown to reduce lung cancer mortality and is recommended for individuals meeting age and smoking history criteria. Despite the expansion of lung cancer screening guidelines in 2021, racial/ethnic and sex disparities persist. High-risk racial minorities and women are more likely to be diagnosed with lung cancer at younger ages and have lower smoking histories when compared with White and male counterparts, resulting in higher rates of ineligibility for screening. Risk prediction models, biomarkers, and deep learning may help refine the selection of individuals who would benefit from screening.

Biomarkers of lung cancer for screening and in never-smokers-a narrative review

Background and Objective

Lung cancer is the leading cause of cancer-related mortality worldwide, partially attributed to late-stage diagnoses. In order to mitigate this, lung cancer screening (LCS) of high-risk patients is performed using low dose computed tomography (CT) scans, however this method is burdened by high false-positive rates and radiation exposure for patients. Further, screening programs focus on individuals with heavy smoking histories, and as such, never-smokers who may otherwise be at risk of lung cancer are often overlooked. To resolve these limitations, biomarkers have been posited as potential supplements or replacements to low-dose CT, and as such, a large body of research in this area has been produced. However, comparatively little information exists on their clinical efficacy and how this compares to current LCS strategies.

Methods

Here we conduct a search and narrative review of current literature surrounding biomarkers of lung cancer to supplement LCS, and biomarkers of lung cancer in never-smokers (LCINS).

Key Content and Findings

Many potential biomarkers of lung cancer have been identified with varying levels of sensitivity, specificity, clinical efficacy, and supporting evidence. Of the markers identified, multi-target panels of circulating microRNAs, lipids, and metabolites are likely the most clinically efficacious markers to aid current screening programs, as these provide the highest sensitivity and specificity for lung cancer detection. However, circulating lipid and metabolite levels are known to vary in numerous systemic pathologies, highlighting the need for further validation in large cohort randomised studies.

Conclusions

Lung cancer biomarkers is a fast-expanding area of research and numerous biomarkers with potential clinical applications have been identified. However, in all cases the level of evidence supporting clinical efficacy is not yet at a level at which it can be translated to clinical practice. The priority now should be to validate existing candidate markers in appropriate clinical contexts and work to integrating these into clinical practice.

Association between metabolically healthy obesity and metastasis in lung cancer patients - a systematic review and meta-analysis

Background

Many clinical trials have looked at the relationship between obesity and lung cancer (LC), however, there is scarcity of literature specifically addressing the association between metabolically healthy obesity and metastasis in LC patients. To address this gap in the body of evidence, the study was conducted to observe the association between metabolically healthy obesity and metastasis in LC patients.

Methods

We conducted a pre-registered systematic review by searching six major online databases to identify studies relevant related to our investigation, in adherence with the PRISMA guidelines. A proper data extraction protocol was further established to synthesize the findings from the selected papers through a meta-analysis.

Results

Eleven (11) studies met the requisite selection criterion and were included in the study. A random-effect model was used. Obesity was found to have a significant impact on readmission in LC patients. The combined analysis showed a significant effect size of 0.08 (95% CI 0.07 to 0.08), indicating a noticeable impact of obesity. It was also assessed that obese individuals had a 34% reduced risk of LC compared to normal weight individuals. Obesity was associated with a lower risk of surgical complications with a pooled risk ratio of 0.13 (95% CI 0.12 to 0.14). A statistically significant decreased risk of LC (pooled RR = 0.72, 95% CI: 0.68 to 0.77) was also observed in the obese individuals.

Conclusion

The analysis reveals that obesity is associated with a noticeable increase in readmissions, although the impact on LC risk itself is negligible. Moreover, obesity appears to have a beneficial effect by reducing the risk of surgical complications. These results highlight the complex relationship between the two aforementioned factors, emphasizing the importance of considering obesity as a significant factor in patient management and healthcare decision-making.

Systematic review registration

https://www.crd.york.ac.uk/prospero/, identifier CRD42023427612.

Assessing eligibility for lung cancer screening using parsimonious ensemble machine learning models: A development and validation study

BACKGROUND

Risk-based screening for lung cancer is currently being considered in several countries; however, the optimal approach to determine eligibility remains unclear. Ensemble machine learning could support the development of highly parsimonious prediction models that maintain the performance of more complex models while maximising simplicity and generalisability, supporting the widespread adoption of personalised screening. In this work, we aimed to develop and validate ensemble machine learning models to determine eligibility for risk-based lung cancer screening.

METHODS AND FINDINGS

For model development, we used data from 216,714 ever-smokers recruited between 2006 and 2010 to the UK Biobank prospective cohort and 26,616 high-risk ever-smokers recruited between 2002 and 2004 to the control arm of the US National Lung Screening (NLST) randomised controlled trial. The NLST trial randomised high-risk smokers from 33 US centres with at least a 30 pack-year smoking history and fewer than 15 quit-years to annual CT or chest radiography screening for lung cancer. We externally validated our models among 49,593 participants in the chest radiography arm and all 80,659 ever-smoking participants in the US Prostate, Lung, Colorectal and Ovarian (PLCO) Screening Trial. The PLCO trial, recruiting from 1993 to 2001, analysed the impact of chest radiography or no chest radiography for lung cancer screening. We primarily validated in the PLCO chest radiography arm such that we could benchmark against comparator models developed within the PLCO control arm. Models were developed to predict the risk of 2 outcomes within 5 years from baseline: diagnosis of lung cancer and death from lung cancer. We assessed model discrimination (area under the receiver operating curve, AUC), calibration (calibration curves and expected/observed ratio), overall performance (Brier scores), and net benefit with decision curve analysis. Models predicting lung cancer death (UCL-D) and incidence (UCL-I) using 3 variables-age, smoking duration, and pack-years-achieved or exceeded parity in discrimination, overall performance, and net benefit with comparators currently in use, despite requiring only one-quarter of the predictors. In external validation in the PLCO trial, UCL-D had an AUC of 0.803 (95% CI: 0.783, 0.824) and was well calibrated with an expected/observed (E/O) ratio of 1.05 (95% CI: 0.95, 1.19). UCL-I had an AUC of 0.787 (95% CI: 0.771, 0.802), an E/O ratio of 1.0 (95% CI: 0.92, 1.07). The sensitivity of UCL-D was 85.5% and UCL-I was 83.9%, at 5-year risk thresholds of 0.68% and 1.17%, respectively, 7.9% and 6.2% higher than the USPSTF-2021 criteria at the same specificity. The main limitation of this study is that the models have not been validated outside of UK and US cohorts.

CONCLUSIONS

We present parsimonious ensemble machine learning models to predict the risk of lung cancer in ever-smokers, demonstrating a novel approach that could simplify the implementation of risk-based lung cancer screening in multiple settings.

Results of a pilot risk-based lung cancer screening study: outcomes and comparisons to a Medicare eligible cohort

PURPOSE

Risk-based lung cancer screening holds potential to detect more cancers and avert more cancer deaths than screening based on age and smoking history alone, but has not been widely assessed or implemented in the United States. The purpose of this study was to prospectively identify patients for lung cancer screening based on lung cancer risk using the PLCOm2012 model and to compare characteristics, risk profiles, and screening outcomes to a traditionally eligible screening cohort.

METHODS

Participants who had a 6 year lung cancer risk score ≥ 1.5% calculated by the PLCOm2012 model and were ineligible for screening under 2015 Medicare guidelines were recruited from a lung cancer screening clinic. After informed consent, participants completed shared decision-making counseling and underwent a low-dose CT (LDCT). Characteristics and screening outcomes of the study population were compared to the traditionally eligible Medicare cohort with Fisher's Exact, t-tests, or Brown Mood tests, as appropriate.

RESULTS

From August 2016 to July 2019, the study completed 48 baseline LDCTs. 10% of LDCTs recommended further pulmonary nodule evaluation (Lung-RADs 3 or 4) with two early-stage lung cancers diagnosed in individuals that had quit smoking > 15 years prior. The study population was approximately 5 years older (p = 0.001) and had lower pack years (p = 0.002) than the Medicare cohort.

CONCLUSION

Prospective application of risk-based screening identifies screening candidates who are similar to a traditionally eligible Medicare cohort and future research should focus on the impact of risk calculators on lung cancer outcomes and optimal usability in clinical environments. This study was retrospectively registered on clinicaltrials.gov (NCT03683940) on 09/25/2018.

The value of blood-based measures of liver function and urate in lung cancer risk prediction: A cohort study and health economic analysis

BACKGROUND

Several studies have reported associations between low-cost blood-based measurements and lung cancer but their role in risk prediction is unclear. We examined the value of expanding lung cancer risk models for targeting low-dose computed tomography (LDCT), including blood measurements of liver function and urate.

METHODS

We analysed a cohort of 388,199 UK Biobank participants with 1873 events and calculated the c-index and fraction of new information (FNI) for models expanded to include combinations of blood measurements, lung function (forced expiratory volume in 1 s - FEV1), alcohol status and waist circumference. We calculated the hypothetical cost per lung cancer case detected by LDCT for different scenarios using a threshold of ≥ 1.51 % risk at 6 years.

RESULTS

The c-index was 0.805 (95 %CI:0.794-0.816) for the model containing conventional predictors. Expanding to include blood measurements increased the c-index to 0.815 (95 %CI: 0.804-0.826;p < 0.0001;FNI:0.06). Expanding to include FEV1, alcohol status, and waist circumference increased the c-index to 0.811 (95 %CI: 0.800-0.822;p < 0.0001;FNI: 0.04). The c-index for the fully expanded model containing all variables was 0.819 (95 %CI:0.808-0.830;p < 0.0001;FNI:0.09). Model expansion had a greater impact on the c-index and FNI for people with a history of smoking cigarettes relative to the full cohort. Compared with the conventional risk model, the expanded models reduced the number of participants meeting the criteria for LDCT screening by 15-21 %, and lung cancer cases detected by 7-8 %. The additional cost per lung cancer case detected relative to the conventional model was £ 1018 for adding blood tests and £ 9775 for the fully expanded model.

CONCLUSION

Blood measurements of liver function and urate made a modest improvement to lung cancer risk prediction compared with a model containing conventional risk factors. There was no evidence that model expansion would improve the cost per lung cancer case detected in UK healthcare settings.

Defining the road map to a UK national lung cancer screening programme

Lung cancer screening with low-dose CT was recommended by the UK National Screening Committee (UKNSC) in September, 2022, on the basis of data from trials showing a reduction in lung cancer mortality. These trials provide sufficient evidence to show clinical efficacy, but further work is needed to prove deliverability in preparation for a national roll-out of the first major targeted screening programme. The UK has been world leading in addressing logistical issues with lung cancer screening through clinical trials, implementation pilots, and the National Health Service (NHS) England Targeted Lung Health Check Programme. In this Policy Review, we describe the consensus reached by a multiprofessional group of experts in lung cancer screening on the key requirements and priorities for effective implementation of a programme. We summarise the output from a round-table meeting of clinicians, behavioural scientists, stakeholder organisations, and representatives from NHS England, the UKNSC, and the four UK nations. This Policy Review will be an important tool in the ongoing expansion and evolution of an already successful programme, and provides a summary of UK expert opinion for consideration by those organising and delivering lung cancer screenings in other countries.

Obesity, the Adipose Organ and Cancer in Humans: Association or Causation?

Epidemiological observations, experimental studies and clinical data show that obesity is associated with a higher risk of developing different types of cancer; however, proof of a cause-effect relationship that meets the causality criteria is still lacking. Several data suggest that the adipose organ could be the protagonist in this crosstalk. In particular, the adipose tissue (AT) alterations occurring in obesity parallel some tumour behaviours, such as their theoretically unlimited expandability, infiltration capacity, angiogenesis regulation, local and systemic inflammation and changes to the immunometabolism and secretome. Moreover, AT and cancer share similar morpho-functional units which regulate tissue expansion: the adiponiche and tumour-niche, respectively. Through direct and indirect interactions involving different cellular types and molecular mechanisms, the obesity-altered adiponiche contributes to cancer development, progression, metastasis and chemoresistance. Moreover, modifications to the gut microbiome and circadian rhythm disruption also play important roles. Clinical studies clearly demonstrate that weight loss is associated with a decreased risk of developing obesity-related cancers, matching the reverse-causality criteria and providing a causality correlation between the two variables. Here, we provide an overview of the methodological, epidemiological and pathophysiological aspects, with a special focus on clinical implications for cancer risk and prognosis and potential therapeutic interventions.

Sybil: A Validated Deep Learning Model to Predict Future Lung Cancer Risk From a Single Low-Dose Chest Computed Tomography

PURPOSE

Low-dose computed tomography (LDCT) for lung cancer screening is effective, although most eligible people are not being screened. Tools that provide personalized future cancer risk assessment could focus approaches toward those most likely to benefit. We hypothesized that a deep learning model assessing the entire volumetric LDCT data could be built to predict individual risk without requiring additional demographic or clinical data.

METHODS

We developed a model called Sybil using LDCTs from the National Lung Screening Trial (NLST). Sybil requires only one LDCT and does not require clinical data or radiologist annotations; it can run in real time in the background on a radiology reading station. Sybil was validated on three independent data sets: a heldout set of 6,282 LDCTs from NLST participants, 8,821 LDCTs from Massachusetts General Hospital (MGH), and 12,280 LDCTs from Chang Gung Memorial Hospital (CGMH, which included people with a range of smoking history including nonsmokers).

RESULTS

Sybil achieved area under the receiver-operator curves for lung cancer prediction at 1 year of 0.92 (95% CI, 0.88 to 0.95) on NLST, 0.86 (95% CI, 0.82 to 0.90) on MGH, and 0.94 (95% CI, 0.91 to 1.00) on CGMH external validation sets. Concordance indices over 6 years were 0.75 (95% CI, 0.72 to 0.78), 0.81 (95% CI, 0.77 to 0.85), and 0.80 (95% CI, 0.75 to 0.86) for NLST, MGH, and CGMH, respectively.

CONCLUSION

Sybil can accurately predict an individual's future lung cancer risk from a single LDCT scan to further enable personalized screening. Future study is required to understand Sybil's clinical applications. Our model and annotations are publicly available.

[Media: see text].

Important parameters for cost-effective implementation of lung cancer screening

It is now widely accepted that lung cancer screening through low-dose computed tomography (LDCT) results in fewer diagnoses at a late stage, and decreased lung cancer mortality. Whilst reducing deaths from lung cancer is an essential prerequisite, this must be balanced against the considerable economic costs accumulated in screening. Multiple health economic models have shown substantial variation in cost per Quality-Adjusted Life Year (QALY), partly driven by the healthcare costs in the country concerned and partly by other modifiable programme components. Recent modelling using UK costs and a targeted approach suggest that most scenarios are within the willingness to pay threshold for the UK. However, identifying the most clinically and cost-effective programme is a priority to minimise the total financial impact. Programme components that influence cost-effectiveness include the method of selection of the eligible population, the participation rate, the interval between rounds of screening, the method of pulmonary nodule management, and the approach to clinical work up. Future research will clarify if a personalised approach to screening, using baseline and subsequent risk to define screening intervals is more cost-effective. The burden of LDCT screening on the medical infrastructure and workforce has to be quantified and carefully managed during implementation.

Predicting the future risk of lung cancer: development, and internal and external validation of the CanPredict (lung) model in 19·67 million people and evaluation of model performance against seven other risk prediction models

BACKGROUND

Lung cancer is the second most common cancer in incidence and the leading cause of cancer deaths worldwide. Meanwhile, lung cancer screening with low-dose CT can reduce mortality. The UK National Screening Committee recommended targeted lung cancer screening on Sept 29, 2022, and asked for more modelling work to be done to help refine the recommendation. This study aims to develop and validate a risk prediction model-the CanPredict (lung) model-for lung cancer screening in the UK and compare the model performance against seven other risk prediction models.

METHODS

For this retrospective, population-based, cohort study, we used linked electronic health records from two English primary care databases: QResearch (Jan 1, 2005-March 31, 2020) and Clinical Practice Research Datalink (CPRD) Gold (Jan 1, 2004-Jan 1, 2015). The primary study outcome was an incident diagnosis of lung cancer. We used a Cox proportional-hazards model in the derivation cohort (12·99 million individuals aged 25-84 years from the QResearch database) to develop the CanPredict (lung) model in men and women. We used discrimination measures (Harrell's C statistic, D statistic, and the explained variation in time to diagnosis of lung cancer [R²_D]) and calibration plots to evaluate model performance by sex and ethnicity, using data from QResearch (4·14 million people for internal validation) and CPRD (2·54 million for external validation). Seven models for predicting lung cancer risk (Liverpool Lung Project [LLP]_v2, LLP_v3, Lung Cancer Risk Assessment Tool [LCRAT], Prostate, Lung, Colorectal, and Ovarian [PLCO]_M2012, PLCO_M2014, Pittsburgh, and Bach) were selected to compare their model performance with the CanPredict (lung) model using two approaches: (1) in ever-smokers aged 55-74 years (the population recommended for lung cancer screening in the UK), and (2) in the populations for each model determined by that model's eligibility criteria.

FINDINGS

There were 73 380 incident lung cancer cases in the QResearch derivation cohort, 22 838 cases in the QResearch internal validation cohort, and 16 145 cases in the CPRD external validation cohort during follow-up. The predictors in the final model included sociodemographic characteristics (age, sex, ethnicity, Townsend score), lifestyle factors (BMI, smoking and alcohol status), comorbidities, family history of lung cancer, and personal history of other cancers. Some predictors were different between the models for women and men, but model performance was similar between sexes. The CanPredict (lung) model showed excellent discrimination and calibration in both internal and external validation of the full model, by sex and ethnicity. The model explained 65% of the variation in time to diagnosis of lung cancer R²_D in both sexes in the QResearch validation cohort and 59% of the R²_D in both sexes in the CPRD validation cohort. Harrell's C statistics were 0·90 in the QResearch (validation) cohort and 0·87 in the CPRD cohort, and the D statistics were 2·8 in the QResearch (validation) cohort and 2·4 in the CPRD cohort. Compared with seven other lung cancer prediction models, the CanPredict (lung) model had the best performance in discrimination, calibration, and net benefit across three prediction horizons (5, 6, and 10 years) in the two approaches. The CanPredict (lung) model also had higher sensitivity than the current UK recommended models (LLP_v2 and PLCO_M2012), as it identified more lung cancer cases than those models by screening the same amount of individuals at high risk.

INTERPRETATION

The CanPredict (lung) model was developed, and internally and externally validated, using data from 19·67 million people from two English primary care databases. Our model has potential utility for risk stratification of the UK primary care population and selection of individuals at high risk of lung cancer for targeted screening. If our model is recommended to be implemented in primary care, each individual's risk can be calculated using information in the primary care electronic health records, and people at high risk can be identified for the lung cancer screening programme.

FUNDING

Innovate UK (UK Research and Innovation).

TRANSLATION

For the Chinese translation of the abstract see Supplementary Materials section.

Evaluation of the accuracy of the PLCOm2012 6-year lung cancer risk prediction model among smokers in the CARTaGENE population-based cohort

BACKGROUND

The PLCO_m2012 prediction tool for risk of lung cancer has been proposed for a pilot program for lung cancer screening in Quebec, but has not been validated in this population. We sought to validate PLCO_m2012 in a cohort of Quebec residents, and to determine the hypothetical performance of different screening strategies.

METHODS

We included smokers without a history of lung cancer from the population-based CARTaGENE cohort. To assess PLCO_m2012 calibration and discrimination, we determined the ratio of expected to observed number of cases, as well as the sensitivity, specificity and positive predictive values of different risk thresholds. To assess the performance of screening strategies if applied between Jan. 1, 1998, and Dec. 31, 2015, we tested different thresholds of the PLCO_m2012 detection of lung cancer over 6 years (1.51%, 1.70% and 2.00%), the criteria of Quebec's pilot program (for people aged 55-74 yr and 50-74 yr) and recommendations from 2021 United States and 2016 Canada guidelines. We assessed shift and serial scenarios of screening, whereby eligibility was assessed annually or every 6 years, respectively.

RESULTS

Among 11 652 participants, 176 (1.51%) lung cancers were diagnosed in 6 years. The PLCO_m2012 tool underestimated the number of cases (expected-to-observed ratio 0.68, 95% confidence interval [CI] 0.59-0.79), but the discrimination was good (C-statistic 0.727, 95% CI 0.679-0.770). From a threshold of 1.51% to 2.00%, sensitivities ranged from 52.3% (95% CI 44.6%-59.8%) to 44.9% (95% CI 37.4%-52.6%), specificities ranged from 81.6% (95% CI 80.8%-82.3%) to 87.7% (95% CI 87.0%-88.3%) and positive predictive values ranged from 4.2% (95% CI 3.4%-5.1%) to 5.3% (95% CI 4.2%-6.5%). Overall, 8938 participants had sufficient data to test performance of screening strategies. If eligibility was estimated annually, Quebec pilot criteria would have detected fewer cancers than PLCO_m2012 at a 2.00% threshold (48.3% v. 50.2%) for a similar number of scans per detected cancer. If eligibility was estimated every 6 years, up to 26 fewer lung cancers would have been detected; however, this scenario led to higher positive predictive values (highest for PLCO_m2012 with a 2.00% threshold at 6.0%, 95% CI 4.8%-7.3%).

INTERPRETATION

In a cohort of Quebec smokers, the PLCO_m2012 risk prediction tool had good discrimination in detecting lung cancer, but it may be helpful to adjust the intercept to improve calibration. The implementation of risk prediction models in some of the provinces of Canada should be done with caution.

Risk Model-Based Lung Cancer Screening : A Cost-Effectiveness Analysis

BACKGROUND

In their 2021 lung cancer screening recommendation update, the U.S. Preventive Services Task Force (USPSTF) evaluated strategies that select people based on their personal lung cancer risk (risk model-based strategies), highlighting the need for further research on the benefits and harms of risk model-based screening.

OBJECTIVE

To evaluate and compare the cost-effectiveness of risk model-based lung cancer screening strategies versus the USPSTF recommendation and to explore optimal risk thresholds.

DESIGN

Comparative modeling analysis.

DATA SOURCES

National Lung Screening Trial; Surveillance, Epidemiology, and End Results program; U.S. Smoking History Generator.

TARGET POPULATION

1960 U.S. birth cohort.

TIME HORIZON

45 years.

PERSPECTIVE

U.S. health care sector.

INTERVENTION

Annual low-dose computed tomography in risk model-based strategies that start screening at age 50 or 55 years, stop screening at age 80 years, with 6-year risk thresholds between 0.5% and 2.2% using the PLCOm2012 model.

OUTCOME MEASURES

Incremental cost-effectiveness ratio (ICER) and cost-effectiveness efficiency frontier connecting strategies with the highest health benefit at a given cost.

RESULTS OF BASE-CASE ANALYSIS

Risk model-based screening strategies were more cost-effective than the USPSTF recommendation and exclusively comprised the cost-effectiveness efficiency frontier. Among the strategies on the efficiency frontier, those with a 6-year risk threshold of 1.2% or greater were cost-effective with an ICER less than $100 000 per quality-adjusted life-year (QALY). Specifically, the strategy with a 1.2% risk threshold had an ICER of $94 659 (model range, $72 639 to $156 774), yielding more QALYs for less cost than the USPSTF recommendation, while having a similar level of screening coverage (person ever-screened 21.7% vs. USPSTF's 22.6%).

RESULTS OF SENSITIVITY ANALYSES

Risk model-based strategies were robustly more cost-effective than the 2021 USPSTF recommendation under varying modeling assumptions.

LIMITATION

Risk models were restricted to age, sex, and smoking-related risk predictors.

CONCLUSION

Risk model-based screening is more cost-effective than the USPSTF recommendation, thus warranting further consideration.

PRIMARY FUNDING SOURCE

National Cancer Institute (NCI).

Lung cancer screening

Randomised controlled trials, including the National Lung Screening Trial (NLST) and the NELSON trial, have shown reduced mortality with lung cancer screening with low-dose CT compared with chest radiography or no screening. Although research has provided clarity on key issues of lung cancer screening, uncertainty remains about aspects that might be critical to optimise clinical effectiveness and cost-effectiveness. This Review brings together current evidence on lung cancer screening, including an overview of clinical trials, considerations regarding the identification of individuals who benefit from lung cancer screening, management of screen-detected findings, smoking cessation interventions, cost-effectiveness, the role of artificial intelligence and biomarkers, and current challenges, solutions, and opportunities surrounding the implementation of lung cancer screening programmes from an international perspective. Further research into risk models for patient selection, personalised screening intervals, novel biomarkers, integrated cardiovascular disease and chronic obstructive pulmonary disease assessments, smoking cessation interventions, and artificial intelligence for lung nodule detection and risk stratification are key opportunities to increase the efficiency of lung cancer screening and ensure equity of access.

Immediate, Remote Smoking Cessation Intervention in Participants Undergoing a Targeted Lung Health Check: Quit Smoking Lung Health Intervention Trial, a Randomized Controlled Trial

BACKGROUND

Lung cancer screening programs provide an opportunity to support people who smoke to quit, but the most appropriate model for delivery remains to be determined. Immediate face-to-face smoking cessation support for people undergoing screening can increase quit rates, but it is not known whether remote delivery of immediate smoking cessation counselling and pharmacotherapy in this context also is effective.

RESEARCH QUESTION

Does an immediate telephone smoking cessation intervention increase quit rates compared with usual care among a population enrolled in a targeted lung health check (TLHC)?

STUDY DESIGN AND METHODS

In a single-masked randomized controlled trial, people 55 to 75 years of age who smoke and attended a TLHC were allocated by day of attendance to receive either immediate telephone smoking cessation intervention (TSI) support (starting immediately and lasting for 6 weeks) with appropriate pharmacotherapy or usual care (UC; very brief advice to quit and signposting to smoking cessation services). The primary outcome was self-reported 7-day point prevalence smoking abstinence at 3 months. Differences between groups were assessed using logistic regression.

RESULTS

Three hundred fifteen people taking part in the screening program who reported current smoking with a mean ± SD age of 63 ± 5.4 years, 48% of whom were women, were randomized to TSI (n = 152) or UC (n = 163). The two groups were well matched at baseline. Self-reported quit rates were higher in the intervention arm, 21.1% vs 8.9% (OR, 2.83; 95% CI, 1.44-5.61; P = .002). Controlling for participant demographics, neither baseline smoking characteristics nor the discovery of abnormalities on low-dose CT imaging modified the effect of the intervention.

INTERPRETATION

Immediate provision of an intensive telephone-based smoking cessation intervention including pharmacotherapy, delivered within a targeted lung screening context, is associated with increased smoking abstinence at 3 months.

TRIAL REGISTRY

ISRCTN registry; No.: ISRCTN12455871; URL: www.IRSCN.com.

OWL: an optimized and independently validated machine learning prediction model for lung cancer screening based on the UK Biobank, PLCO, and NLST populations

BACKGROUND

A reliable risk prediction model is critically important for identifying individuals with high risk of developing lung cancer as candidates for low-dose chest computed tomography (LDCT) screening. Leveraging a cutting-edge machine learning technique that accommodates a wide list of questionnaire-based predictors, we sought to optimize and validate a lung cancer prediction model.

METHODS

We developed an Optimized early Warning model for Lung cancer risk (OWL) using the XGBoost algorithm with 323,344 participants from the England area in UK Biobank (training set), and independently validated it with 93,227 participants from UKB Scotland and Wales area (validation set 1), as well as 70,605 and 66,231 participants in the Prostate, Lung, Colorectal, and Ovarian cancer screening trial (PLCO) control and intervention subpopulations, respectively (validation sets 2 & 3) and 23,138 and 18,669 participants in the United States National Lung Screening Trial (NLST) control and intervention subpopulations, respectively (validation sets 4 & 5). By comparing with three competitive prediction models, i.e., PLCO modified 2012 (PLCO_m2012), PLCO modified 2014 (PLCO_all2014), and the Liverpool Lung cancer Project risk model version 3 (LLPv3), we assessed the discrimination of OWL by the area under receiver operating characteristic curve (AUC) at the designed time point. We further evaluated the calibration using relative improvement in the ratio of expected to observed lung cancer cases (RI_EO), and illustrated the clinical utility by the decision curve analysis.

FINDINGS

For general population, with validation set 1, OWL (AUC = 0.855, 95% CI: 0.829-0.880) presented a better discriminative capability than PLCO_all2014 (AUC = 0.821, 95% CI: 0.794-0.848) (p < 0.001); with validation sets 2 & 3, AUC of OWL was comparable to PLCO_all2014 (AUC_PLCOall2014-AUC_OWL < 1%). For ever-smokers, OWL outperformed PLCO_m2012 and PLCO_all2014 among ever-smokers in validation set 1 (AUC_OWL = 0.842, 95% CI: 0.814-0.871; AUC_PLCOm2012 = 0.792, 95% CI: 0.760-0.823; AUC_PLCOall2014 = 0.791, 95% CI: 0.760-0.822, all p < 0.001). OWL remained comparable to PLCO_m2012 and PLCO_all2014 in discrimination (AUC difference from -0.014 to 0.008) among the ever-smokers in validation sets 2 to 5. In all the validation sets, OWL outperformed LLPv3 among the general population and the ever-smokers. Of note, OWL showed significantly better calibration than PLCO_m2012, PLCO_all2014 (RI_EO from 43.1% to 92.3%, all p < 0.001), and LLPv3 (RI_EO from 41.4% to 98.7%, all p < 0.001) in most cases. For clinical utility, OWL exhibited significant improvement in average net benefits (NB) over PLCO_all2014 in validation set 1 (NB improvement: 32, p < 0.001); among ever smokers of validation set 1, OWL (average NB = 289) retained significant improvement over PLCO_m2012 (average NB = 213) (p < 0.001). OWL had equivalent NBs with PLCO_m2012 and PLCO_all2014 in PLCO and NLST populations, while outperforming LLPv3 in the three populations.

INTERPRETATION

OWL, with a high degree of predictive accuracy and robustness, is a general framework with scientific justifications and clinical utility that can aid in screening individuals with high risks of lung cancer.

FUNDING

National Natural Science Foundation of China, the US NIH.

[China National Lung Cancer Screening Guideline with Low-dose Computed Tomography (2023 Version)]

Lung cancer is the leading cause of cancer-related death in China. The effectiveness of low-dose computed tomography (LDCT) screening has been further validated in recent years, and significant progress has been made in research on identifying high-risk individuals, personalizing screening interval, and management of screen-detected findings. The aim of this study is to revise China national lung cancer screening guideline with LDCT (2018 version). The China Lung Cancer Early Detection and Treatment Expert Group (CLCEDTEG) designated by the China's National Health Commission, and China Lung Oncolgy Group experts, jointly participated in the revision of Chinese lung cancer screening guideline (2023 version). This revision is based on the recent advances in LDCT lung cancer screening at home and abroad, and the epidemiology of lung cancer in China. The following aspects of the guideline were revised: (1) lung cancer risk factors besides smoking were considered for the identification of high risk population; (2) LDCT scan parameters were further classified; (3) longer screening interval is recommended for individuals who had negative LDCT screening results for two consecutive rounds; (4) the follow-up interval for positive nodules was extended from 3 months to 6 months; (5) the role of multi-disciplinary treatment (MDT) in the management of positive nodules, diagnosis and treatment of lung cancer were emphasized. This revision clarifies the screening, intervention and treatment pathways, making the LDCT screening guideline more appropriate for China. Future researches based on emerging technologies, including biomarkers and artificial intelligence, are needed to optimize LDCT screening in China in the future.
.

Comparing risk prediction models aimed at predicting hospitalizations for adverse drug events in community dwelling older adults: a protocol paper

Background

The objective of this paper is to describe the creation, validation, and comparison of two risk prediction modeling approaches for community-dwelling older adults to identify individuals at highest risk for adverse drug event-related hospitalizations. One approach will use traditional statistical methods, the second will use a machine learning approach.

Methods

We will construct medication, clinical, health care utilization, and other variables known to be associated with adverse drug event-related hospitalizations. To create the cohort, we will include older adults (≥ 65 years of age) empaneled to a primary care physician within the Cedars-Sinai Health System primary care clinics with polypharmacy (≥ 5 medications) or at least 1 medication commonly implicated in ADEs (certain oral hypoglycemics, anti-coagulants, anti-platelets, and insulins). We will use a Fine-Gray Cox proportional hazards model for one risk modeling approach and DataRobot, a data science and analytics platform, to run and compare several widely used supervised machine learning algorithms, including Random Forest, Support Vector Machine, Extreme Gradient Boosting (XGBoost), Decision Tree, Naïve Bayes, and K-Nearest Neighbors. We will use a variety of metrics to compare model performance and to assess the risk of algorithmic bias.

Discussion

In conclusion, we hope to develop a pragmatic model that can be implemented in the primary care setting to risk stratify older adults to further optimize medication management.

A systematic review of genetic ancestry as a risk factor for incidence of non-small cell lung cancer in the US

Background: Non-Small Cell Lung Cancer (NSCLC), the leading cause of cancer-related death in the United States, is the most diagnosed form of lung cancer. While lung cancer incidence has steadily declined over the last decade, disparities in incidence and mortality rates persist among African American (AA), Caucasian American (CA), and Hispanic American (HA) populations. Researchers continue to explore how genetic ancestry may influence differential outcomes in lung cancer risk and development. The purpose of this evaluation is to highlight experimental research that investigates the differential impact of genetic mutations and ancestry on NSCLC incidence. Methods: This systematic review was conducted using PubMed and Google Scholar search engines. The following key search terms were used to select articles published between 2011 and 2022: "African/European/Latin American Ancestry NSCLC"; "Racial Disparities NSCLC"; "Genetic Mutations NSCLC"; "NSCLC Biomarkers"; "African Americans/Hispanic Americans/Caucasian Americans NSCLC incidence." Systematic reviews, meta-analyses, and studies outside of the US were excluded. A total of 195 articles were initially identified and after excluding 156 which did not meet eligibility criteria, 38 were included in this investigation. Results: Studies included in this analysis focused on racial/ethnic disparities in the following common genetic mutations observed in NSCLC: KRAS, EGFR, TP53, PIK3CA, ALK Translocations, ROS-1 Rearrangements, STK11, MET, and BRAF. Results across studies varied with respect to absolute differential expression. No significant differences in frequencies of specific genetic mutational profiles were noted between racial/ethnic groups. However, for HAs, lower mutational frequencies in KRAS and STK11 genes were observed. In genetic ancestry level analyses, multiple studies suggest that African ancestry is associated with a higher frequency of EGFR mutations. Conversely, Latin ancestry is associated with TP53 mutations. At the genomic level, several novel predisposing variants associated with African ancestry and increased risk of NSCLC were discovered. Family history among all racial/ethnic groups was also considered a risk factor for NSCLC. Conclusion: Results from racially and ethnically diverse studies can elucidate driving factors that may increase susceptibility and subsequent lung cancer risk across different racial/ethnic groups. Identification of biomarkers that can be used as diagnostic, prognostic, and therapeutic tools may help improve lung cancer survival among high-risk populations.

Development and Validation of the Random Forest Model via Combining CT-PET Image Features and Demographic Data for Distant Metastases among Lung Cancer Patients

The work aimed at developing and validating a random forest model of CT-PET image features combined with demographic data to diagnose distant metastases among lung cancer patients. This study involved lung cancer patients from The Cancer Genome Atlas lung adenocarcinoma (TCGA-LUAD) dataset, the lung PET-CT dataset, the lung squamous cell carcinoma (LSCC) dataset, and the National Cancer Institute's Clinical Proteomic Tumor Analysis Consortium lung adenocarcinoma (CPTAC-LUAD) dataset and collected the information on 178 CT, 178 PET, and the patients' age, history of smoking, and gender. We conducted image processing and feature extraction. Finally, 4 computed tomography (CT) image features and 2 positron emission tomography (PET) image features were extracted. Four prediction models based on CT image features, PET image features, and demographic data were developed, and the area under the receiver operating characteristic (ROC) curve was used to evaluate the performance of prediction models. A total of 178 eligible samples were randomly divided into a training set (n = 134) and a testing set (n = 44) at a ratio of 3 : 1, with 2021 as a random number. ROC analyses illustrated that the predictive performance for distant metastases of combining CT-PET image features and demographic data for training and testing were 0.923 (95% confidence interval (CI): 0.873-0.973) and 0.873 (95% CI: 0.757-0.990). In addition, the predictive performance of the combined model in the testing set was significantly better than that of the CT-demographic data model (0.716, 95% CI: 0.531-0.902), PET-demographic data model (0.802, 95% CI: 0.633-0.970), and CT-PET model (0.797, 95% CI: 0.666-0.928). The random forest model via combining CT-PET image features and demographic data could have great performance in predicting distant metastases among lung cancer patients.

Recalibrating Risk Prediction Models by Synthesizing Data Sources: Adapting the Lung Cancer PLCO Model for Taiwan

BACKGROUND

Methods synthesizing multiple data sources without prospective datasets have been proposed for absolute risk model development. This study proposed methods for adapting risk models for another population without prospective cohorts, which would help alleviate the health disparities caused by advances in absolute risk models. To exemplify, we adapted the lung cancer risk model PLCOM2012, well studied in the west, for Taiwan.

METHODS

Using Taiwanese multiple data sources, we formed an age-matched case-control study of ever-smokers (AMCCSE), estimated the number of ever-smoking lung cancer patients in 2011-2016 (NESLP2011), and synthesized a dataset resembling the population of cancer-free ever-smokers in 2010 regarding the PLCOM2012 risk factors (SPES2010). The AMCCSE was used to estimate the overall calibration slope, and the requirement that NESLP2011 equals the estimated total risk of individuals in SPES2010 was used to handle the calibration-in-the-large problem.

RESULTS

The adapted model PLCOT-1 (PLCOT-2) had an AUC of 0.78 (0.75). They had high performance in calibration and clinical usefulness on subgroups of SPES2010 defined by age and smoking experience. Selecting the same number of individuals for low-dose computed tomography screening using PLCOT-1 (PLCOT-2) would have identified approximately 6% (8%) more lung cancers than the US Preventive Services Task Forces 2021 criteria. Smokers having 40+ pack-years had an average PLCOT-1 (PLCOT-2) risk of 3.8% (2.6%).

CONCLUSIONS

The adapted PLCOT models had high predictive performance.

IMPACT

The PLCOT models could be used to design lung cancer screening programs in Taiwan. The methods could be applicable to other cancer models.

Comparison of discrimination performance of 11 lung cancer risk models for predicting lung cancer in a prospective cohort of screening-age adults from Germany followed over 17 years

Randomized trials have demonstrated considerable reduction in lung cancer (LC) mortality by screening pre-selected heavy smokers with low-dose computed tomography (LDCT). Newer screening guidelines recommend refined LC risk models for selecting the target population for screening. We aimed to evaluate and compare the discrimination performance of LC risk models and previously used trial criteria in predicting LC incidence and mortality in a large German cohort of screening-age adults. Within ESTHER, a population-based prospective cohort study conducted in Saarland, Germany, 4812 ever smokers aged 50-75 years were followed up with respect to LC incidence and mortality for up to 17 years. We quantified the performance of 11 different LC risk models by the area under the curve (AUC) and compared the proportion of correctly predicted LC cases between the best performing models and the LDCT trial criteria. Risk prediction of LC incidence in the ESTHER ever smokers was best for the Bach model, LCRAT and LCDRAT with AUCs ranging from 0.782 to 0.787, from 0.770 to 0.774, and from 0.765 to 0.771 for the follow-up time periods of cases identified at 6, 11, and 17 years, respectively. At cutoffs yielding comparable positivity rates as the LDCT trial criteria, these models would have identified between 11.8 (95% CI 3.0-20.5) and 17.6 (95% CI 10.1-25.2) percent units higher proportions of LC cases occurring during the initial 6 years of follow-up. Use of LC risk models is expected to result in substantially greater potential to identify people at highest risk of LC, suggesting enhanced potential for reducing LC mortality by LC screening.

Non-Invasive Biomarkers for Early Lung Cancer Detection

Worldwide, lung cancer (LC) is the most common cause of cancer death, and any delay in the detection of new and relapsed disease serves as a major factor for a significant proportion of LC morbidity and mortality. Though invasive methods such as tissue biopsy are considered the gold standard for diagnosis and disease monitoring, they have several limitations. Therefore, there is an urgent need to identify and validate non-invasive biomarkers for the early diagnosis, prognosis, and treatment of lung cancer for improved patient management. Despite recent progress in the identification of non-invasive biomarkers, currently, there is a shortage of reliable and accessible biomarkers demonstrating high sensitivity and specificity for LC detection. In this review, we aim to cover the latest developments in the field, including the utility of biomarkers that are currently used in LC screening and diagnosis. We comment on their limitations and summarise the findings and developmental stages of potential molecular contenders such as microRNAs, circulating tumour DNA, and methylation markers. Furthermore, we summarise research challenges in the development of biomarkers used for screening purposes and the potential clinical applications of newly discovered biomarkers.

Cost-Effectiveness Analyses of Lung Cancer Screening Using Low-Dose Computed Tomography: A Systematic Review Assessing Strategy Comparison and Risk Stratification

OBJECTIVES

Our first study objective was to assess the range of lung cancer screening intervals compared within cost-effectiveness analyses (CEAs) of low-dose computed tomography (LDCT) and to examine the implications for the strategies identified as optimally cost effective; the second objective was to examine if and how risk subgroup-specific policies were considered.

METHODS

PubMed, Embase and Web of Science were searched for model-based CEAs of LDCT lung screening. The retrieved studies were assessed to examine if the analyses considered sufficient strategy variation to permit incremental estimation of cost effectiveness. Regarding risk selection, we examined if analyses considered alternative risk strata in separate analyses or as alternative risk-based eligibility criteria for screening.

RESULTS

The search identified 33 eligible CEAs, 23 of which only considered one screening frequency. Of the 10 analyses considering multiple screening intervals, only 4 included intervals longer than 2 years. Within the 10 studies considering multiple intervals, the optimal policy choice would differ in 5 if biennial intervals or longer had not been considered. Nineteen studies conducted risk subgroup analyses, 12 of which assumed that subgroup-specific policies were possible and 7 of which assumed that a common screening policy applies to all those screened.

CONCLUSIONS

The comparison of multiple strategies is recognised as good practice in CEA when seeking optimal policies. Studies that do include multiple intervals indicate that screening intervals longer than 1 year can be relevant. The omission of intervals of 2 years or longer from CEAs of LDCT screening could lead to the adoption of sub-optimal policies. There also is scope for greater consideration of risk-stratified policies which tailor screening intensity to estimated disease risk. Policy makers should take care when interpreting current evidence before implementing lung screening.

From the patient to the population: Use of genomics for population screening

Genomic medicine is expanding from a focus on diagnosis at the patient level to prevention at the population level given the ongoing under-ascertainment of high-risk and actionable genetic conditions using current strategies, particularly hereditary breast and ovarian cancer (HBOC), Lynch Syndrome (LS) and familial hypercholesterolemia (FH). The availability of large-scale next-generation sequencing strategies and preventive options for these conditions makes it increasingly feasible to screen pre-symptomatic individuals through public health-based approaches, rather than restricting testing to high-risk groups. This raises anew, and with urgency, questions about the limits of screening as well as the moral authority and capacity to screen for genetic conditions at a population level. We aimed to answer some of these critical questions by using the WHO Wilson and Jungner criteria to guide a synthesis of current evidence on population genomic screening for HBOC, LS, and FH.

The need for a comprehensive and integrated approach to lung cancer policy in Europe

Lung cancer is the leading cause of cancer-related deaths in Europe. Europe's Beating Cancer Plan calls for a comprehensive approach to the disease in general but not specifically to lung cancer. Such a comprehensive approach, integrating efforts to strengthen anti-tobacco policies, early detection and underlying models of care, is sorely needed for lung cancer - particularly considering disruptions to care during the COVID-19 pandemic. In a recently published think piece, a multidisciplinary group of experts proposed four key policy priority areas. First, to reduce stigma and improve awareness of potential symptoms, there is a need to foster a better understanding of lung cancer - among the public and healthcare professionals. Second, opportunities for early detection should be enhanced, and the implementation of targeted screening through low-dose computed tomography should be encouraged as a complement to smoking cessation services. This complementarity should be recognised and built into joint policy proposals, with development and better integration of screening and smoking cessation programmes on the ground. Third, the socioeconomic inequalities underpinning disparities in outcomes in people with lung cancer must be addressed, with targeted approaches to overcome barriers to access Finally, the overall quality of lung cancer care must be improved, making multidisciplinary care available to all and ensuring survivorship is given due attention.

The Journey for Lung Cancer Screening where we Stand Today

Background

Lung cancer remains a leading cause of cancer mortality worldwide with many patients presenting with advanced disease.

Objective

We reviewed the available literature for lung cancer screening using low dose computed tomography (LDCT). We reviewed the National Lung Screening Trial (NLST), Early Lung Cancer Action Program (ELCAP) and the (Nederlands-Leuvens Longkanker Screenings Onderzoek (NELSON) trials. We also look at different lung cancer risk prediction models that may aid in identifying target populations and also discuss the cost-effectiveness of LDCT screening in different groups of smokers and ex-smokers. Lastly, we discuss recent guideline changes that have occurred in line with new and emerging evidence on lung cancer screening.

Conclusion

LDCT has been shown reduce lung cancer mortality in certain groups of current and former smokers and should be considered to help in the early diagnosis of lung cancer.

Subpopulation-specific machine learning prognosis for underrepresented patients with double prioritized bias correction

Background

Many clinical datasets are intrinsically imbalanced, dominated by overwhelming majority groups. Off-the-shelf machine learning models that optimize the prognosis of majority patient types (e.g., healthy class) may cause substantial errors on the minority prediction class (e.g., disease class) and demographic subgroups (e.g., Black or young patients). In the typical one-machine-learning-model-fits-all paradigm, racial and age disparities are likely to exist, but unreported. In addition, some widely used whole-population metrics give misleading results.

Methods

We design a double prioritized (DP) bias correction technique to mitigate representational biases in machine learning-based prognosis. Our method trains customized machine learning models for specific ethnicity or age groups, a substantial departure from the one-model-predicts-all convention. We compare with other sampling and reweighting techniques in mortality and cancer survivability prediction tasks.

Results

We first provide empirical evidence showing various prediction deficiencies in a typical machine learning setting without bias correction. For example, missed death cases are 3.14 times higher than missed survival cases for mortality prediction. Then, we show DP consistently boosts the minority class recall for underrepresented groups, by up to 38.0%. DP also reduces relative disparities across race and age groups, e.g., up to 88.0% better than the 8 existing sampling solutions in terms of the relative disparity of minority class recall. Cross-race and cross-age-group evaluation also suggests the need for subpopulation-specific machine learning models.

Conclusions

Biases exist in the widely accepted one-machine-learning-model-fits-all-population approach. We invent a bias correction method that produces specialized machine learning prognostication models for underrepresented racial and age groups. This technique may reduce potentially life-threatening prediction mistakes for minority populations.

Derivation and validation of a prediction model for patients with lung nodules malignancy regardless of mediastinal/hilar lymphadenopathy

BACKGROUND

Clinical prediction models to classify lung nodules often exclude patients with mediastinal/hilar lymphadenopathy, although the presence of mediastinal/hilar lymphadenopathy does not always indicate malignancy. Herein, we developed and validated a multimodal prediction model for lung nodules in which patients with mediastinal/hilar lymphadenopathy were included.

METHODS

A single-center retrospective study was conducted. We developed and validated a logistic regression model including patients with mediastinal/hilar lymphadenopathy. Discrimination of the model was assessed by area under the operating curve. Goodness of fit test was performed via the Hosmer-Lemeshow test, and a nomogram of the logistic regression model was drawn.

RESULTS

There were 311 cases included in the final analysis. A logistic regression model was developed and validated. There were nine independent variables included in the model. The aera under the curve (AUC) of the validation set was 0.91 (95% confidence interval [CI]: 0.85-0.98). In the validation set with mediastinal/hilar lymphadenopathy, the AUC was 0.95 (95% CI: 0.90-0.99). The goodness-of-fit test was 0.22.

CONCLUSIONS

We developed and validated a multimodal risk prediction model for lung nodules with excellent discrimination and calibration, regardless of mediastinal/hilar lymphadenopathy. This broadens the application of lung nodule prediction models. Furthermore, mediastinal/hilar lymphadenopathy added value for predicting lung nodule malignancy in clinical practice.

Impact of low-dose computed tomography (LDCT) screening on lung cancer-related mortality

BACKGROUND

Lung cancer is the most common cause of cancer-related death in the world, however lung cancer screening has not been implemented in most countries at a population level. A previous Cochrane Review found limited evidence for the effectiveness of lung cancer screening with chest radiography (CXR) or sputum cytology in reducing lung cancer-related mortality, however there has been increasing evidence supporting screening with low-dose computed tomography (LDCT).  OBJECTIVES: To determine whether screening for lung cancer using LDCT of the chest reduces lung cancer-related mortality and to evaluate the possible harms of LDCT screening.

SEARCH METHODS

We performed the search in collaboration with the Information Specialist of the Cochrane Lung Cancer Group and included the Cochrane Lung Cancer Group Trial Register, Cochrane Central Register of Controlled Trials (CENTRAL, the Cochrane Library, current issue), MEDLINE (accessed via PubMed) and Embase in our search. We also searched the clinical trial registries to identify unpublished and ongoing trials. We did not impose any restriction on language of publication. The search was performed up to 31 July 2021.  SELECTION CRITERIA: Randomised controlled trials (RCTs) of lung cancer screening using LDCT and reporting mortality or harm outcomes.  DATA COLLECTION AND ANALYSIS: Two review authors were involved in independently assessing trials for eligibility, extraction of trial data and characteristics, and assessing risk of bias of the included trials using the Cochrane RoB 1 tool. We assessed the certainty of evidence using GRADE. Primary outcomes were lung cancer-related mortality and harms of screening. We performed a meta-analysis, where appropriate, for all outcomes using a random-effects model. We only included trials in the analysis of mortality outcomes if they had at least 5 years of follow-up. We reported risk ratios (RRs) and hazard ratios (HRs), with 95% confidence intervals (CIs) and used the I2 statistic to investigate heterogeneity.  MAIN RESULTS: We included 11 trials in this review with a total of 94,445 participants. Trials were conducted in Europe and the USA in people aged 40 years or older, with most trials having an entry requirement of ≥ 20 pack-year smoking history (e.g. 1 pack of cigarettes/day for 20 years or 2 packs/day for 10 years etc.). One trial included male participants only. Eight trials were phase three RCTs, with two feasibility RCTs and one pilot RCT. Seven of the included trials had no screening as a comparison, and four trials had CXR screening as a comparator. Screening frequency included annual, biennial and incrementing intervals. The duration of screening ranged from 1 year to 10 years. Mortality follow-up was from 5 years to approximately 12 years.  None of the included trials were at low risk of bias across all domains. The certainty of evidence was moderate to low across different outcomes, as assessed by GRADE. In the meta-analysis of trials assessing lung cancer-related mortality, we included eight trials (91,122 participants), and there was a reduction in mortality of 21% with LDCT screening compared to control groups of no screening or CXR screening (RR 0.79, 95% CI 0.72 to 0.87; 8 trials, 91,122 participants; moderate-certainty evidence). There were probably no differences in subgroups for analyses by control type, sex, geographical region, and nodule management algorithm. Females appeared to have a larger lung cancer-related mortality benefit compared to males with LDCT screening. There was also a reduction in all-cause mortality (including lung cancer-related) of 5% (RR 0.95, 95% CI 0.91 to 0.99; 8 trials, 91,107 participants; moderate-certainty evidence).  Invasive tests occurred more frequently in the LDCT group (RR 2.60, 95% CI 2.41 to 2.80; 3 trials, 60,003 participants; moderate-certainty evidence). However, analysis of 60-day postoperative mortality was not significant between groups (RR 0.68, 95% CI 0.24 to 1.94; 2 trials, 409 participants; moderate-certainty evidence).  False-positive results and recall rates were higher with LDCT screening compared to screening with CXR, however there was low-certainty evidence in the meta-analyses due to heterogeneity and risk of bias concerns. Estimated overdiagnosis with LDCT screening was 18%, however the 95% CI was 0 to 36% (risk difference (RD) 0.18, 95% CI -0.00 to 0.36; 5 trials, 28,656 participants; low-certainty evidence). Four trials compared different aspects of health-related quality of life (HRQoL) using various measures. Anxiety was pooled from three trials, with participants in LDCT screening reporting lower anxiety scores than in the control group (standardised mean difference (SMD) -0.43, 95% CI -0.59 to -0.27; 3 trials, 8153 participants; low-certainty evidence). There were insufficient data to comment on the impact of LDCT screening on smoking behaviour.  AUTHORS' CONCLUSIONS: The current evidence supports a reduction in lung cancer-related mortality with the use of LDCT for lung cancer screening in high-risk populations (those over the age of 40 with a significant smoking exposure). However, there are limited data on harms and further trials are required to determine participant selection and optimal frequency and duration of screening, with potential for significant overdiagnosis of lung cancer. Trials are ongoing for lung cancer screening in non-smokers.

Screening Criteria Evaluation for Expansion in Pulmonary Neoplasias (SCREEN)

The SCREEN study investigated screening eligibility and survival outcomes between heavy smokers and light-or-never-smokers with lung cancer to determine whether expanded risk factor analysis is needed to refine screening criteria. SCREEN is a retrospective study of 917 lung cancer patients diagnosed between 2005 and 2018 in Nova Scotia, Canada. Screening eligibility was determined using the National Lung Screening Trial (NSLT) criteria. Mortality risk between heavy smokers and light-or-never-smokers was compared using proportional-hazards models. The median follow-up was 2.9 years. The cohort was comprised of 179 (46.1%) female heavy smokers and 306 (57.8%) female light-or-never-smokers. Light-or-never-smokers were more likely to have a diagnosis of adenocarcinoma [n=378 (71.6%)] compared to heavy smokers [n=234 (60.5%); P< 0.001]. Heavy smokers were more frequently diagnosed with squamous cell carcinoma [n=111 (28.7%)] compared to light-or-never-smokers, [n=100 (18.9%); P< 0.001]. Overall, 36.9% (338) of patients met NLST screening criteria. There was no difference in 5-year survival between light-or-never-smokers and heavy smokers [55.2% (338) vs 58.5% (529); P = 0.408; HR 1.06, 95% CI 0.80-1.40; P = 0.704]. Multivariate analysis showed that males had an increased mortality risk [HR 2.00 (95% CI 1.57-2.54); P< 0.001]. Half of lung cancer patients were missed with the conventional screening criteria. There were more curable, stage 1 tumors among light-or-never-smokers. Smoking status and age alone may be insufficient predictors of lung cancer risk and prognosis. Expanded risk factor analysis is needed to refine lung cancer screening criteria.