Prognosis prediction is central in treatment decision making and quality of life for non-small cell lung cancer (NSCLC) patients. However, conventional computed tomography (CT) related prognostic factors may not apply to the challenging stage III NSCLC group. The aim of this systematic review was therefore to identify and evaluate CT-related prognostic factors for overall survival (OS) of stage III NSCLC. The Medline, Embase, and Cochrane electronic databases were searched. After study selection, risk of bias was estimated for the included studies. Meta-analysis of univariate results was performed when sufficient data were available. 1595 of the 11,996 retrieved records were selected for full text review, leading to inclusion of 65 studies that reported data of 144,513 stage III NSCLC patients andcompromising 26 unique CT-related prognostic factors. Relevance and validity varied substantially, few studies had low relevance and validity. Only four studies evaluated the added value of new prognostic factors compared with recognized clinical factors. Included studies suggested gross tumor volume (meta-analysis: HR = 1.22, 95%CI: 1.05-1.42), tumor diameter, nodal volume, and pleural effusion, are prognostic in patients treated with chemoradiation. Clinical T-stage and location (right/left) were likely not prognostic within stage III NSCLC. Inconclusive are several radiomic features, tumor volume, atelectasis, location (pulmonary lobes, central/peripheral), interstitial lung abnormalities, great vessel invasion, pit-fall sign, and cavitation. Tumor-size and nodal size-related factors are prognostic for OS in stage III NSCLC. Future studies should carefully report study characteristics and contrast factors with guideline recognized factors to improve evidence evaluation and validation.

Clinicians increasingly rely on prediction models to guide treatment choices. Most prediction models, however, are developed using observational data that include some patients who have already received the treatment the prediction model is meant to inform. Special attention to the causal role of those earlier treatments is required when interpreting the resulting predictions. "Causal blind spots" were identified in 3 common approaches to handling treatment when developing a prediction model: including treatment as a predictor, restricting to persons taking a certain treatment, and ignoring treatment. Through several real examples, this article illustrates how the risks obtained from models developed using such approaches may be misinterpreted and can lead to misinformed decision making. The discussion covers issues attributable to confounding, selection, mediation, and changes in treatment protocols over time. An extension of guidelines for the development, reporting, and evaluation of prediction models is advocated to avoid such misinterpretations. Developers must ensure that the intended target population for the model, and the treatment conditions under which predictions hold, are clearly communicated. When prediction models are intended to inform treatment decisions, they need to provide estimates of risk under the specific treatment (or intervention) options being considered, known as "prediction under interventions." Next to suitable data, this requires causal reasoning and causal inference techniques during model development and evaluation. Being clear about what a given prediction model can and cannot be used for prevents misinformed treatment decisions and thereby prevents potential harm to patients.

Deep learning has shown remarkable results for image analysis and is expected to aid individual treatment decisions in health care. Treatment recommendations are predictions with an inherently causal interpretation. To use deep learning for these applications in the setting of observational data, deep learning methods must be made compatible with the required causal assumptions. We present a scenario with real-world medical images (CT-scans of lung cancer) and simulated outcome data. Through the data simulation scheme, the images contain two distinct factors of variation that are associated with survival, but represent a collider (tumor size) and a prognostic factor (tumor heterogeneity), respectively. When a deep network would use all the information available in the image to predict survival, it would condition on the collider and thereby introduce bias in the estimation of the treatment effect. We show that when this collider can be quantified, unbiased individual prognosis predictions are attainable with deep learning. This is achieved by (1) setting a dual task for the network to predict both the outcome and the collider and (2) enforcing a form of linear independence of the activation distributions of the last layer. Our method provides an example of combining deep learning and structural causal models to achieve unbiased individual prognosis predictions. Extensions of machine learning methods for applications to causal questions are required to attain the long-standing goal of personalized medicine supported by artificial intelligence.

Oncologists are faced with choosing the best treatment for each patient, based on the available evidence from randomized controlled trials (RCTs) and observational studies. RCTs provide estimates of the average effects of treatments on groups of patients, but they may not apply in many real-world scenarios where for example patients have different characteristics than the RCT participants, or where different treatment variants are considered. Causal inference defines what a treatment effect is and how it may be estimated with RCTs or outside of RCTs with observational - or 'real-world' - data. In this review, we introduce the field of causal inference, explain what a treatment effect is and what important challenges are with treatment effect estimation with observational data. We then provide a framework for conducting causal inference studies and describe when in oncology causal inference from observational data may be particularly valuable. Recognizing the strengths and limitations of both RCTs and observational causal inference provides a way for more informed and individualized treatment decision-making in oncology.

The presence of tumor-infiltrating lymphocytes (TILs) in melanoma has been linked to survival. Their predictive capability for immune checkpoint inhibition (ICI) response remains uncertain. Therefore, we investigated the association between treatment response and TILs in the largest cohort to date and analyzed if this association was independent of known clinical predictors. In this multicenter cohort study, patients who received first-line anti-PD1 ± anti-CTLA4 for advanced melanoma were identified. TILs were scored on hematoxylin and eosin (H&E) slides of primary melanoma and pre-treatment metastases using the validated TILs-WG, Clark and MIA score. The primary outcome was objective response rate (ORR), with progression free survival and overall survival being secondary outcomes. Univariable and multivariable logistic regression and Cox proportional hazard were performed, adjusting for known clinical predictors. Metastatic melanoma specimens were available for 650 patients and primary specimens for 565 patients. No association was found in primary melanoma specimens. In metastatic specimens, a 10-point increase in the TILs-WG score was associated with a higher probability of response (aOR 1.17, 95 % CI 1.07-1.28), increased PFS (HR 0.93, 95 % CI 0.87-0.996), and OS (HR 0.94, 95 % CI 0.89-0.99). When categorized, patients in the highest tertile TILs-WG score (15-100 %) compared to the lowest tertile (0 %) had a longer median PFS (13.1 vs. 7.3 months, p = 0.04) and OS (49.4 vs. 19.5 months, p = 0.003). Similar results were noted using the MIA and Clark scores. In advanced melanoma patients, TIL patterns on H&E slides of pre-treatment metastases, regardless of measurement method, are independently associated with ICI response. TILs are easily scored on readily available H&Es, facilitating the use of this biomarker in clinical practice.