Brain imaging in lung cancer patients without symptoms of brain metastases: a national survey of current practice in England

Machine Learning for the Prediction of Synchronous Organ-Specific Metastasis in Patients With Lung Cancer

Background

This study aimed to develop an artificial neural network (ANN) model for predicting synchronous organ-specific metastasis in lung cancer (LC) patients.

Methods

A total of 62,151 patients who diagnosed as LC without data missing between 2010 and 2015 were identified from Surveillance, Epidemiology, and End Results (SEER) program. The ANN model was trained and tested on an 75/25 split of the dataset. The receiver operating characteristic (ROC) curves, area under the curve (AUC) and sensitivity were used to evaluate and compare the ANN model with the random forest model.

Results

For distant metastasis in the whole cohort, the ANN model had metrics AUC = 0.759, accuracy = 0.669, sensitivity = 0.906, and specificity = 0.613, which was better than the random forest model. For organ-specific metastasis in the cohort with distant metastasis, the sensitivity in bone metastasis, brain metastasis and liver metastasis were 0.913, 0.906 and 0.925, respectively. The most important variable was separate tumor nodules with 100% importance. The second important variable was visceral pleural invasion for distant metastasis, while histology for organ-specific metastasis.

Conclusions

Our study developed a “two-step” ANN model for predicting synchronous organ-specific metastasis in LC patients. This ANN model may provide clinicians with more personalized clinical decisions, contribute to rationalize metastasis screening, and reduce the burden on patients and the health care system.

Overuse of Diagnostic Brain Imaging Among Patients With Stage IA Non-Small Cell Lung Cancer

BACKGROUND

Among patients diagnosed with stage IA non-small cell lung cancer (NSCLC), the incidence of occult brain metastasis is low, and several professional societies recommend against brain imaging for staging purposes. The goal of this study was to characterize the use of brain imaging among Medicare patients diagnosed with stage IA NSCLC.

METHODS

Using data from linked SEER-Medicare claims, we identified patients diagnosed with AJCC 8th edition stage IA NSCLC in 2004 through 2013. Patients were classified as having received brain imaging if they underwent head CT or brain MRI from 1 month before to 3 months after diagnosis. We identified factors associated with receipt of brain imaging using multivariable logistic regression.

RESULTS

Among 13,809 patients with stage IA NSCLC, 3,417 (25%) underwent brain imaging at time of diagnosis. The rate of brain imaging increased over time, from 23.5% in 2004 to 28.7% in 2013 (P=.0006). There was significant variation in the use of brain imaging across hospital service areas, with rates ranging from 0% to 64.0%. Factors associated with a greater likelihood of brain imaging included older age (odds ratios [ORs] of 1.16 for 70-74 years, 1.13 for 75-79 years, 1.31 for 80-84 years, and 1.46 for ≥85 years compared with 65-69 years; all P<.05), female sex (OR, 1.09; P<.05), black race (OR 1.23; P<.05), larger tumor size (ORs of 1.23 for 11-20 mm and 1.28 for 21-30 mm tumors vs 1-10 mm tumors; all P<.05), and higher modified Charlson-Deyo comorbidity score (OR, 1.28 for score >1 vs score of 0; P<.05).

CONCLUSIONS

Roughly 1 in 4 patients with stage IA NSCLC received brain imaging at the time of diagnosis despite national recommendations against the practice. Although several patient factors are associated with receipt of brain imaging, there is significant geographic variation across the United States. Closer adherence to clinical guidelines is likely to result in more cost-effective care.

Effects of different brain surveillance strategies on outcomes for patients with EGFR-mutant metastatic lung adenocarcinoma under targeted therapy

OBJECTIVES

Brain metastasis (BM) is common in patients with epidermal growth factor receptor (EGFR)-mutant lung cancer. However, the brain surveillance strategy during treatment in advanced lung cancer patients varies, and the impact on clinical outcome is unclear. Here we aimed to evaluate the effect of different brain surveillance strategies on the clinical characteristics and treatment outcome in patients with EGFR-mutant lung adenocarcinoma treated with first-line EGFR tyrosine kinase inhibitors (EGFR-TKIs).

MATERIALS AND METHODS

This is a retrospective observational study conducted in a medical center in an area with high prevalence of EGFR mutation. Patients with initially diagnosed stage IV EGFR-mutant lung adenocarcinoma were included. Patients undergoing regular brain magnetic resonance imaging (MRI) every 3-6 months were categorized in the regular follow-up (RFU) group, and the rest were categorized in the liberal follow-up (LFU) group. Clinical outcomes were collected and analyzed.

RESULTS

A total of 310 patients were included, and 43.5% initially had brain metastases. Patients in the LFU group were significantly older than those in the RFU group (median age: 67 vs 62, p < 0.001). The overall survival and time-to-treatment failure of patients with initial EGFR-TKIs treatment showed no statistical difference between the two groups. However, the intracranial progression free survival was significantly shorter in the RFU group than in the LFU group (p = 0.009). The risk of mortality was similar in the LFU and RFU groups. There was no difference in the intracranial progression patterns and cause of death between the two groups.

CONCLUSIONS

For EGFR-mutant lung adenocarcinoma patients who used EGFR-TKIs as the frontline therapy, regular or liberal brain MRI follow-up showed no significant impact on the outcome, irrespective of initial brain metastasis.

Screening for brain metastases in patients with stage III non-small-cell lung cancer, magnetic resonance imaging or computed tomography? A prospective study

INTRODUCTION

Non-small-cell lung cancer (NSCLC) guidelines advise to screen stage III NSCLC patients for brain metastases (BMs), preferably by magnetic resonance imaging (MRI) or when contraindicated or not accessible a dedicated contrast enhanced-computed tomography (dCE-CT), which can be incorporated in the staging ¹⁸Fluodeoxoglucose-positron emission tomography (¹⁸FDG-PET-CE-CT). In daily practice, often a dCE-CT is performed instead of a MRI. The aim of the current study is to evaluate the additive value of MRI after dCE-CT, incorporated in the ¹⁸FDG-PET-CE-CT.

PATIENTS AND METHODS

It is an observational prospective multicentre study (NTR3628). Inclusion criteria included stage III NSCLC patients with a dCE-CT of the brain incorporated in the ¹⁸FDG-PET and an additional MRI of the brain. Primary end-point is percentage of patients with BM on MRI without suspect lesions on dCE-CT. Secondary end-points are percentage of patients with BM on dCE-CT and percentage of patients with BM ≤ 1 year of a negative staging MRI.

RESULTS

Sixteen (7%) patients with extracranial stage III had BM on dCE-CT and were excluded. One hundred forty-nine patients were enrolled. 7/149 (4.7%) had BM on MRI without suspect lesions on dCE-CT. One hundred eighteen patients had a follow-up of at least 1 year (four with BM on baseline MRI); eight of the remaining 114 (7%) patients developed BM ≤ 1 year after a negative staging brain MRI.

CONCLUSION

Although in 7% of otherwise stage III NSCLC patients, BMs were detected on staging dCE-CT, MRI brain detected BMs in an additional 4.7%, which we consider clinically relevant. Within 1 year after a negative staging MRI, 7% developed BM.

What you see is (not) what you get: tools for a non-radiologist to evaluate image quality in lung cancer

Medical images are an integral part of oncological patient records and they are reviewed by many different specialists. Therefore, it is important that besides imaging experts, other clinicians are also aware that the diagnostic value of a scan is influenced by the applied imaging protocol. Based on two clinical lung cancer trials, we experienced that, even within a study protocol, there is a large variability in imaging parameters, which has direct impact on the interpretation of the image. These two trials were: 1) the NTR3628 in which the added value of gadolinium magnetic resonance imaging (Gd-MRI) to dedicated contrast enhanced computed tomography (CE-CT) for detecting asymptomatic brain metastases in stage III non-small cell lung cancer (NSCLC) was investigated and 2) a sub-study of the NVALT 12 trial (NCT01171170) in which repeated 18 F-fludeoxyglucose positron emission tomography (¹⁸F-FDG-PET) imaging for early response assessment was investigated. Based on the problems encountered in the two trials, we provide recommendations for non-radiology clinicians, which can be used in daily interpretation of imaging. Variations in image parameters cannot only influence trial results, but sub-optimal imaging can also influence treatment decisions in daily lung cancer care, when a physician is not aware of the scanning details.

Diversity of brain metastases screening and management in non-small cell lung cancer in Europe: Results of the European Organisation for Research and Treatment of Cancer Lung Cancer Group survey

BACKGROUND

Brain metastases (BM) are frequent in non-small cell lung cancer (NSCLC) patients, but there is a lack of evidence-based management of this patient group. We aimed to capture a snapshot of routine BM management in Europe to identify relevant research questions for future clinical trials.

METHODS

An EORTC Lung Cancer Group (LCG) online survey containing questions on NSCLC BM screening and treatment was distributed between 16/02/17 and 15/06/17 to worldwide EORTC LCG members, and through several European scientific societies in the thoracic oncology field.

RESULTS

A total of 462 European physician responses (394 institutions) were analysed (radiation oncologist: 53% [n = 247], pulmonologist: 26% [n = 119], medical oncologist: 18% [n = 84]; 84% with >5 years' experience in NSCLC). Italy (18%, n = 85), Netherlands (15%, n = 68), UK (14%, n = 66), and France (12%, n = 55) contributed most. 393 physicians (85%) screened neurologically asymptomatic patients for BM at diagnosis (52% using magnetic resonance imaging). Most often screened patients were those with a driver mutation (MUT+; 51%, n = 234), stage III (63%, n = 289), and IV (43%, n = 199). 158 physicians (34%) used a prognostic classification to guide initial treatment decisions, and in 50%, lowest prognostic-score threshold to receive treatment differed between MUT+ and non-driver mutation (MUT-) patients. MUT+ patients with >4 BM were more likely to receive stereotactic radiosurgery (SRS) compared with MUT- (27% versus. 21%; p < 0.01). Most physicians (90%) had access to SRS. After single BM surgery, 50% systematically prescribed SRS or WBRT, and 45% only in case of incomplete resection. The preferred treatment in neurologically asymptomatic treatment-naive patients diagnosed with >5 BM was systemic treatment (79%). Of all, 45%/49% physicians stated that all tyrosine kinase inhibitors and immune checkpoint blockers were discontinued (timing varied) during SRS/WBRT, respectively. Drugs most often continued during SRS/WBRT were erlotinib (44%/40%), gefitinib (39%/34%), afatinib (29%/25%), crizotinib (33%/26%) and anti-PD-(L)-1 (28%/22%).

CONCLUSION

BM management is highly variable in Europe: screening is not uniform, prognostic classifications are not often used and MUT+ NSCLC patients generally receive more intensive local treatment. Prospective assessment of BM management in MUT+ NSCLC patients is required.

Lung cancer guidelines in Sweden, Denmark, Norway and Finland: a comparison

BACKGROUND

The Nordic countries are similar in terms of demographics and health care organization. Yet there are marked differences in lung cancer mortality, for which Denmark historically has had the poorest outcome. One of several possible reasons for these differences could have to do with how lung cancer is diagnosed and treated in the different Nordic countries. However, among the four most populous Nordic countries: Sweden, Denmark, Norway and Finland, there is a paucity of knowledge about differences and similarities in recommendations in the national guidelines for non-small cell lung cancer (NSCLC) and the methodology by which the guidelines are developed.

METHODS

We identified and evaluated the development and content of the available clinical care guidelines for NSCLC in the four countries. Moreover, we compared the integrated cancer pathways in these countries. We have used case examples to illustrate areas with clear differences in clinical care recommendations.

RESULTS

There are notable differences in the methodology by which the guidelines are developed, published and updated to comply with international recommendations. The Norwegian guidelines are developed and updated according to the most rigorous methodology and have so far been updated most frequently. We found that on the basis of recommendations patients with NSCLC are treated differently with regard to bevacizumab therapy and radiation dosing regimens. Cerebral imaging practices in patients with locally advanced NSCLC also differ. There is, moreover, a marked difference with regard to efforts to help patients to quit smoking. All except Finland have integrated cancer pathways for fast track diagnosis and treatment. Guidelines for follow-up of lung cancer patients also differ, with the Danish follow-up regimen as the most comprehensive. To obtain consensus on optimal clinical care, areas with differences in recommendations or where recommendations are based on a low level of evidence should be subjected to further studies.

Brain imaging before primary lung cancer resection: a controversial topic

Objective

International and national recommendations for brain imaging in patients planned to undergo potentially curative resection of non-small-cell lung cancer (NSCLC) are variably implemented throughout the United Kingdom [Hudson BJ, Crawford MB, and Curtin J et al (2015) Brain imaging in lung cancer patients without symptoms of brain metastases: a national survey of current practice in EnglandClin Radiol https://doi.org/10.1016/j.crad.2015.02.007]. However, the recommendations are not based on high-quality evidence and do not take into account cost implications and local resources. Our aim was to determine local practice based on historic outcomes in this patient cohort.

Methods

This retrospective study took place in a regional thoracic surgical centre in the United Kingdom. Pathology records for all patients who had undergone lung resection with curative intent during the time period January 2012–December 2014 were analysed in October 2015. Electronic pathology and radiology reports were accessed for each patient and data collected about their histological findings, TNM stage, resection margins, and the presence of brain metastases on either pre-operative or post-operative imaging. From the dates given on imaging, we calculated the number of days post-resection that the brain metastases were detected.

Results

585 patients were identified who had undergone resection of their lung cancer. Of these, 471 had accessible electronic radiology records to assess for the radiological evidence of brain metastases. When their electronic records were evaluated, 25/471 (5.3%) patients had radiological evidence of brain metastasis. Of these, five patients had been diagnosed with a brain metastasis at initial presentation and had undergone primary resection of the brain metastasis followed by resection of the lung primary. One patient had been diagnosed with both a primary lung and a primary bowel adenocarcinoma; on review of the case, it was felt that the brain metastasis was more likely to have originated from the bowel cancer. One had been clinically diagnosed with a cerebral abscess while the radiology had been reported as showing a metastatic deposit. Of the remaining 18/471 (3.8%) patients who presented with brain metastases after their surgical resection, 12 patients had adenocarcinoma, four patients had squamous cell carcinoma, one had basaloid, and one had large-cell neuroendocrine. The mean number of days post-resection that the brain metastases were identified was 371 days, range 14–1032 days, median 295 days (date of metastases not available for two patients).

Conclusion

The rate of brain metastases identified in this study was similar to previous studies. This would suggest that preoperative staging of the central nervous system may change the management pathway in a small group of patients. However, for this group of patients, the change would be significant either sparing them non-curative surgery or allowing aggressive management of oligometastatic disease. Therefore, we would recommend pre-operative brain imaging with MRI for all patients undergoing potentially curative lung resection.

Computed tomography protocols used in staging bronchopulmonary carcinoma: results of a national survey

OBJECTIVES

To know the protocols used for staging bronchopulmonary carcinoma by computed tomography in Spain.

MATERIAL AND METHODS

Radiologists in 129 hospitals were sent email questionnaires about the organization of their department, scanner type and manufacturer, study extension, techniques employed, and protocol for administering contrast material.

RESULTS

A total of 109 hospitals responded with data from 91 teams. Most hospitals were affiliated with a university, and most departments were organized by organ-systems. Scanners were from four manufacturers, and 68% had either 16 or 64 detectors. In 61% of the hospitals, the dose of contrast agent is modified only in patients with extreme body weights, and in 22% the dose is not individualized. Most hospitals do contrast-enhanced studies of the chest and upper abdomen, 42.4% through a single thoracoabdominal acquisition and 55.9% through independent chest and abdominal acquisitions; there was a significant association between these approaches and the scanner manufacturer's protocols and whether the hospital was affiliated with a university. The most commonly used technical parameters were 120kV with dose modulation and variable milliamperage.

CONCLUSION

There is very little variability among hospitals in the type of scanner used, the study extension, and the technical parameters used to stage bronchopulmonary carcinoma. Most centers individualize the dose of contrast agent only in extreme weights. There is a broad division between using one or two acquisitions to image the thorax and abdomen, and the number of acquisitions is related to the scanner manufacturer and whether the hospital is affiliated with a university.